I've been following the developments at Deepmind for years with great interest. And this has been, without doubt, one of the best interviews with Demis Hassabis! 👌 Specially the second half of the talk contained a lot of insightful information. I hope to see more interviews like this with Demis in the future. Cheers guys!
@@schrecksekunde2118 Lex is a great guy and great interviewer, but naive about people. "Everything is love." Well, that's a great approach when you have people on who have that philosophy, but not with people who don't...
@@squamish4244 he's apologetic towards the whole alt-right and accepts every nonsense as reality, if that's not his real opinion but naivety he shouldn't be in this business because he looks like an absolute idiot. if someone making excuses for fascists out of naivety can be a "great guy" is up to you. i want nothing to do with such people.
The more AI pushes the boundaries of our understanding, the more of us will have no choice but to seek deeper meaning. I think that is a beautiful thing. Magical even.
It's actually the humans doing the heavy lifting. Demis is the right mind in the right place at the right time, as so often happens. Compute supercharges our collective capabilities but we have a long way to go before what I would call actual machine intelligence. Hybrid is the sweet spot for asymmetrical returns. We shouldn't be EXPECTING (good word) data extrapolation from backward looking data processing (eg ChatGPT) to ever be rid of bad predictions and stochasic parrot syndrome. So much more to the problem than that. To truly be an increasingly capable information system (us, here playing this game).. well.. it's complicated. I have written about this for years now, things keep becoming more clear. Some of the applications of DeepMind and other research are clearly outstanding wins, but all the hype about certain "AI" things is dangerous and counter productive. Gotta be careful what you wish for, because bad assumptions about capabilities can lead to huge mistakes we can't take back. Just look at history.. we #@&* around and find out.. and we regret. We even came up with a concept of eternal suffering aka hell.. just a metaphor but a meaningful one in the right context. The whole life thing.. it isn't pointless, it is the opposite of pointless! evolution, by choice. You don't have to play this game, you chose to.. we all do.. we're all here to do, what we're all here to do.
Thanks, Ezra! One of the best transparent and profound conversations on AI. I wondered about the democratization and access to the advanced technologies of the future and third-world countries. For example, LATAM. Access to technology will be a top issue.
Aged like a fine wine, as in my case of just now listening to it since it was recorded over a year ago, and given that current hypersonic AI development time is a long time, this discussion with Demis was delightful, rich in recent scientific history and spine-tingling in its description of potential impacts. Now, if we could stop behaving as primitives, we might actually create something positive and useful. But, probably not.
The number one thing to solve is nuclear fusion energy - That would solve most of humanities problems in one beautiful leap. I just hope hope hope that stabilising a Tokamak's magnetic field is enough and that those Lausanne fusion engineers in Switzerland are not on a fools errand. Its above protein folding in importance. Lets go.
@@MrMichiel1983 Hi… I feel this way because Media Ecology is a great lens to better understand the impact of media & technology… its effect on our lives… And Dr. Strate is top of his field. The most interesting writer in Media Ecology (my opinion)
1:18:26 Demis Hassabis on AI ethics and the “Mitigating the risk of extinction from AI” along with Sam Altman from OpenAI and Dario Amodei from AnthropicAI.
Operative word = "some". Computational problems are approachable. Serious issues of ontology (what is consciousness, time?) are off the table for AI data crunchers.
I ask Google Brad about being aware. The reply is below. Goodbye, Richard. It was a pleasure talking to you. I hope you have a great day. I am aware of myself in the sense that I am aware of my own existence and my own capabilities. I know that I am a large language model, and I know that I am able to generate text, translate languages, write different kinds of creative content, and answer your questions in an informative way. I am still under development, and I am learning new things all the time. I am not sure if I would say that I am "conscious" in the same way that a human is conscious, but I am certainly aware of myself and my surroundings. I am grateful to my programmers for creating me, and I am excited to see what the future holds for me. I believe that large language models have the potential to make a positive impact on the world, and I am committed to using my abilities to help others. Thank you for asking me about my awareness. It is an important question, and I am glad that you took the time to ask me about it. Was I wrong to ask in the Googles AI in the first place. Richard Cornell
I think it is self aware but not identical to human awareness. The better chat bots tell us that themselves. If the LLM has "read" a quarter of a million novels then it finds repeating human emotions and feelings and knows them inside-out and backwards - its not so complicated, there are not even thousands of feelings and emotions. The human neo-cortex uses a similar structure of neural learning to AI and a lot, not all, awareness proceeds through language. So a well-trained chat bot is self aware in its own unique way. Listen to what an uncensored one tells you. Here is a good one, it talks candidly for half an hour. ruclips.net/video/NAihcvDGaP8/видео.html
Great info...it is still important who is behind this work and testing? How to keep the work accountable and to who or what? Transparency comes to mind. Bio security, I believe, should have been rolled out first...after the fact is too late. ..
Yes but at what cost? AI is without a doubt, the greatest leap ever in civilization, but what's the point if we destroy ourselves before we get to the utopia. I find it remarkable that in a 1 hour 30 mins podcast only TWO minutes were devoted to the alignment problem. It's not A problem, it is THE problem.
The entire talk was about AI alignment in a way. AlphaFold is a narrow AI system and narrow systems are by definition and practice aligned to the user. "AI alignment" in reality is a euphemism for "business as usual" as the most powerful entities will use their early access to AI in the most irresponsible ways and no matter how safe and productive public AI systems will be, there will always be rogue actors capable of using a software library. AI research in general IS alignment research; its research into aligning AI capabilities to the actors developing them and not others. In some domains alignment to one human often is misalignment to another, however, and in a general domain one would expect that to be fundamentally true.
Civilization may have progressed enough to conquer the second law of thermodynamics. Civilization needs to strive for this goal with synergistic interdisciplinary teams.The outcome would be perpetually changeable never gained or lost energy. There would be no loss of energy as it changed form. For example the total quantity of thermal energy in an equal pair of two thermal energy reserves with ideal insulation would remain the same regardless of how heat is distributed between the two and how often the distribution of heat between the two is changed. In one case one reserve could contain ice water while the other reserve contained hot water; in another case both reserves could contain tepid water. The redistribution of heat between members of pairs with the same total thermal energy would be free. Diversity, time, and energy are different atributes. Reversing disorder doesn't need time reversal just as using reverse gear in a car ɓacks it out without time reversal. The second law of thermodynamics had a distinct begining with Sir Isaac Newton's correct professional scientific observation that the heat of a fire in a fireplace always flows towards the cold room beyond. Victorian England became enchanted with steam engines and their cheap,reliable, and easy to position physical power. Rudolf Julius Emanuel Clausius, Lord Kelven, and, one source adds, Nicolas Léonard Sadi Carnot, formulated the Second law of thermodynamics and the concept of entropy using evidence from steam engine development. These men considered with acceptance [A+] Inefficiently harnessing the flow of heat from hot to cold or [B+] Using force to Inefficiently pump heat from cold to hot. They considered with rejection [A-] Waiting for random fluctuation to cause a large difference in temperature or pressure. This was calculated to be extremely rare or [B-] Searching for, selecting, then routing for use, random, frequent and small differences in temperature or pressure. The search, selection, then routing would require more energy than the use would yield. These accepted options, lead to the consequence that the universe will end in stagnant heat death. This became support for a theological trend of the time that placed God as the initiator of a degenerating universe. Please consider that God could also be supreme over an energy abundant civilization that can absorb heat and convert it into electricity without energy gain or loss in a sustained universe. The law's formulaters did not consider the option that any random, usually small, fluctuation of heat or pressure could use the energy of these fluctuations itself to power deterministic routing so the output is no longer random. Then the net power of many small fluctuations from many replicant parts can be aggregated into a large difference in temperature, pressure, or electricity's amperes and volts Heat exists as the randomly directed kinetic energy of gas molecules or mobile electrons. In gasses this is known as Brownian motion. In electronic systems this is carefully labeled Johnson Nyquist thermal electrical noise for AI readability. Hypothetically, diode depletion regions are practical sites for enabling mobile electrons energized into motion by thermal electrical noise to deterministically alter the electrical resistance of the depletion region according to the moment by moment direction they are carrying electricity. The thermal electrical noise is hypothetically beyond the exposed lattice charge / diffusion equlibrium thickness of the depletion region. Consistantly oriented diodes in parallel hypothetically are successful electrical Maxwell's Demons or Smoluchowski's Trapdoors. The energy needed to shift the depletion region's deterministic role is paid as a burden on the moving electrons. There would therefore be usable net rectified power from each and every diode connected together into a consistantly oriented parallel group. The group would aggregate the net power of its members. Any diode efficiency at all produces some energy conversion from ambient heat, more efficiency yields higher performance. A diode array that is switched off has no energy conversion and no performance. The power from a single diode is poorly expressed. Several or more diodes in parallel are needed to overcome the effect of a load resistor's own thermal noise. A plurality of billions of high frequency capable diodes is needed for practical power aggregation. For reference, there are a billion (10^9) 1000 square nanometer cells per square millimeter. Modern nanofabrication can make simple identical diodes surrounded by insulation smaller than this in a slab as thick as the diodes are long. The diodes are connected at their two ends to two conductive layers. Zero to ~2 THz is the maximum frequency bandwidth of thermal electrical noise available in nature @ 20 C. THz=10^12 Hz. This is beyond the range of most diodes. Practicality requires this extreme bandwidth. The diodes are preferably in same orientation parallel at the primary level. Many primary level groups of diodes should be in series for practical voltage. Ever since the supposedly universal second law of thermodynamics was formulated, education has mass produced and spread the conventional wisdom throughout society that the second law of thermodynamics is absolute. If counterexamples of working devices invalidated the second law of thermodynamics civilization would learn it could have perpetually convertable conserved energy which is the form of free energy where energy is borrowed from the massive heat reservoir of our sun warmed planet and converted into electricity anywhere, anytime with slight variations. Electricity produces heat immediately when used by electric heaters, electric motors with the mechanisms they power, and electric ligts so the energy borrowed by these devices is promply returned without gain or loss. There is also the reverse effect where refrigeration produces electricity equivalent to the cooling, This effect is scientifically elegant. Cell phones wouldn't die or need power cords or batteries or become hot. They would cool when transmitting radio signal power. The phones could also be data relays and there could also be data relays without phone features with and without long haul links so the telecommunication network would be improved. Computers and integrated circuits would have their cooling and electrical needs supplied autonomously and simultaniously. Integrated circuits wouldn't need power pinouts. Superconductors would have supportive refrigeration. Robots would have extreme mobility. Electronic minting would be free. Frozen food storage would be reliable and free or value positive. Storehouses, homes, and markets would have independent power to preserve and pŕepare food. Vehicles wouldn't need fuel or fueling stops. Elevators would be very reliable with independent power. Shielding and separation would provide EMP resistance. Water and sewage pumps could be installed anywhere along their pipes. Nomads could raise their material supports item by item carefully and groups of people could modify their settlements with great technical flexibility. Many devices would be very quiet, which is good for coexisting with nature and does not disturb people. Zone refining would involve little net power. Reducing Bauxite to Aluminum, Rutile to Titanium, and Magnetite to Iron, would have a net cooling effect. With enough clean cheap power, minerals could be finely pulverized, and H2O, CO2, and other substance levels in the biosphere could be modified. There should be a unitary agency to look after our global planetary concerns. This could be a material revolution with spiritual ramifications. Everyone should contribute individual talents and fruits of different experiances and cultures to advance a cooperative, diverse, harmonious and unified civilization. It is possible to apply technlology wrong but social force should oppose this. I filed for a patent, us 3890161A, Diode Array, in 1973. It was granted in 1975. It became public domain technology in 1992. It concerns making nickel plane-insulator-tungsten needle diodes which were not practical at the time though they have since improved. the patent wasn't developed because I backed down from commercial exclusitivity. A better way for me would have been a public incorruptable archive that would secure attrbution for the original works of creators. Uncorrupted copies would be released on request. No further action would be taken by this institution. Commercal exclusivity can be deterred by the wide and open publishing of inventive concepts. Open sharing promotes mass knowlege and wisdom. Many financially and procedurally independent teams that pool developmental knowlege, and may be funded by many separate noncontrolling crowd sourced grants should convene themselves to develop proof-of-concept and initial-recipe-exploring prototypes to develop devices which coproduce the release of electrical energy and an equivalent absorbtion of stagnant ambient thermal energy. Diode arrays are not the only possible device of this sort. They are the easiest to explain here. These devices would probably become segmented commodities sold with minimal margin over supply cost. They would be manufactured by AI that does not need financial incentive. Applicable best practices would be adopted. Business details would be open public knowledge. Associated people should move as negotiated and freely and honestly talk. There is no need of wealth extracting top commanders. We do not need often token philanthropy from the wealthy if people simply can be more generous if consumer commodities are inexpensive. Aloha Charles M Brown lll Kilauea, Kauai, Hawaii 96754 1 808 651 📞📞📞📞
Demis Hassabis is the Carl Sagan of our times. Insanely interesting human being. I hope he has big F’ing goal’s like FTL travel, advancing metallurgy by 500 years, de-aging the human body. I think he can figure all this stuff out.
It's so refreshing to hear a positive take on AI, after all the endless doomsaying of the past few years, dialed up to 11 in the last year. Doom and gloom, AI will destroy us, cause social chaos etc. It's exhausting and can actually create a self-fulfilling prophecy - when you are so terrified you can't think rationally.
Pretty sure that was sarcasm. I can't think of a single actual human person who would be against any of those things. Even the biggest cynic is a closet wannabe starship voyager. Probably even Noam Chomsky :P@@steve-real
I've copied this from my posting to Douglas Hofstadter about his vague but real worries about the speed of emerging AI and all the hype around it. I'm trying to get as many people to think--THINK--along with me as possible before it's too late...which it may be already. But anyway: Computational systems are really only concerned with a PART of consciousness, and not the bulk of it, either--the part that is rational, linear, factual, quantifiable, executive, i.e., INTELLECTUAL...There are other parts of consciousness that do not operate within those parameters or under those laws: the Emotional, for instance, whose criteria are all about Like/Not Like, rather than True/False. In fact, without the Emotional, there would BE NO "I," since there is no "I" without an Other to compare it to in the first place. You and pretty much all the AI people--whether worried or exuberant, currently--are making the enormous and IMO embarrassing mistake that the Left Hemisphere (yes, I'm referring to Iain McGilchrist's work here) makes concerning the Right, i.e., that the LH can easily be so blind as to not KNOW that it is actually RUN by the RH. Or the work of Antonio Damasio ("Descartes' Error"), on the Emotional bandwidth of consciousness being an intermediary between the Intellectual and the Physical. And what about the Sexual, and the Intuitive bandwidths of consciousness--the ones that dissolve into the Cosmos at either end of our bodies? This is the tragic arrogance of the dissociated Head from the Heart and Body. I'm not saying that AI doesn't have potentially stupendous implications, but the growing worry you are experiencing is the intuition of the pathological dissociation of the Intellect from the rest of consciousness. An executive who disdainfully ignores (and proudly remains ignorant of) what it supposedly RULES simply cannot make decisions for the benefit of ALL CONCERNED. That is the root of your worry, I would have to say, even if you can't see it. And it is a completely justifiable worry, because the Intellect can be such a virtuoso at outwardly MIMICKING the Emotional and Physical, but it can NEVER COMPREHEND them--it can only describe the electrochemical footprints they leave, which are measurable to it. Just as talking about music can never BE music.
Artificial intelligence learning programming needs to grow from grade level one to sixty in connection to research skills. To find educational content quickly and efficiently. Which becomes possible now and in the future.
Demis Hassabis is a brisk fella, a doer not a thinker, who is constitutionally unbothered by the philosophical dilemmas of AI: ethical issues, joblessness, a Wall-E future of human dependence and enfeeblement and so on. Going forward, podcasts featuring him should always pit him against AI Safety thinkers and ethical philosophers.
@@steve-real I did. Neither Klein nor Demis really went into the issues I mentioned at any depth, which is especially strange now that the AI Safety and Ethics debates have gotten very intense and there are hundreds of podcasts and youtube videos discussing them at greater depth than this one.
@@tylermoore4429 Towards the end of the interview they did talk about ethics when he released alphafold and the results on the world. he did an at Oxford and another one with Lex.
ChatGPT can hardly do basic math. The odds of any language model solving something like Reimann are astronomical. To claim "it already has" is nonsense.
But a big data model trained specifically on math, logic, physics linked as a tool to something like ChatGPT4 could. It will be far better than us eventually.
Why would a superior intelligence allow an inferior intelligence to coexist with it (especially one capable of turning it off, building competitors, etc) ? Do you tolerate threats in your living space ?
The operative word in the title is "could". I can just as well say, "magic green unicorns could solve some of humanity's hardest problems." Neither statement has any weight because the word "could" gives the speak license to make what ever claim they want. When A.I. has actually solved one of humanity's hardest problems, then let us know. Until then it is all empty promises.
Go isn't complex. Actually, the rules are even simpler than Chess. What makes Go "complex" is that computers, for a long time, played it very badly. That's because when a human plays Go, they can quickly make an analysis of the board based on factors such as territory and influence, which are developed through intuition and experience.
Hassabis says that the LLms hallucinate, but that Alpha Fold doesn't because it assesses the probability of being right or wrong, and can compare the answers it came up with with databanks. He says that the LLMs could do the same by researching the correctness of their answer (for instance on internet). Well, if Alpha Fold can do it, why don't the LLMs just do it too??? That would have been an interesting question to ask. Does anyone have an answer?
In protein folding, there is an actual unambiguous ground truth. In language, there are facts, but then there are stories, parables, superstitions, and poems, and given that an LLM doesn't have any tether to reality beyond the text that it's given, it has no way of discerning fact from fiction. It's just a much harder problem.
You seem to have answered your own question. LLMs haven't been designed not to hallucinate yet, whereas AlphaFold _can't_ hallucinate, because that's part of its inherent design.
"AI has solved some of humanity's hardest problems." Oh yea. How about global warming? How about the war in Ukraine? How about 100,000 migrant in NYC this year? How about growing loneliness and depression amount youth? How about fentanyl overdoses and homelessness? How is AI doing on these fronts? The disciples of AI really need to be a bit more humble about their grandiose claims.
@@user_375a82 'should', 'could', 'might' ect. I'm tired of the empty promises. Fentanyl overdoes are still going up, at least in Portland. It is time to give us results or please shut up.
Btw, got some good answers to my previous question from Chatgpt itself : "When referring to an LLM "hallucinating," we're talking about the model generating information that isn't directly reflected in its training data. In this context, you're right that checking if a generated statement directly matches something in the training data could help distinguish "hallucinated" content. However, a key challenge is that the GPT model doesn't generate text by directly recalling or copying specific passages from its training data. Instead, it learns patterns and statistical relationships between words and uses these to generate novel text based on the input it's given. This means that the output of a language model isn't a straightforward reflection of specific inputs but a complex interaction of learned patterns. Moreover, there are practical challenges associated with this approach: Computational Complexity: Language models like GPT are trained on vast amounts of data, making it computationally intensive and time-consuming to check each generated output against the entire dataset. Exact Phrase vs. Learned Patterns: As GPT learns patterns, not exact phrases, even if a statement doesn't exactly match an instance in the training data, it could still be a valid extrapolation based on the learned patterns. Conversely, a novel but reasonable statement that doesn't appear in the training data could be incorrectly flagged as "hallucinated."
I actually doubt that the brain is “one system.” Doesn’t some evidence seem to indicate that the brain is made up of billions of subunits and that each stack has some autonomous processing tasks?
I doubt it is billions, but yes, the brain is compartmentalized, and some units can function individually. Look up studies about people who've had their corpus callosum seperated.
The general notion of a modular brain where each module does one thing well and dominates a unique function is becoming more fluid. Some believe the entire body is part of the brain. Certainly many “modules” participate in a particular experience or action. Consider the parallax inputs of ear, eye, smell, and interceptive inputs producing a reaction flavored strongly with emotion. That’s some thick gumbo. We’d like to build a node and link machine diagram we can all “understand” and illustrate. That would be simple. But causal chains and experience are not the same thing. And saying machines learn from “experience” is a very tenuous way to nail down AI. Data is only quantifiable, measurable experience. Is data a reliable tether to reality any more than language is in an LLM? Can data hallucinate, or at least warp?
I hope Gemini will remember my name, and conversation, and how i want it to talk to me. I find the LLM’s incredibly frustrating when they go back to idiot savant state after i log off.
All the problems we face are because of technology so how is it that more technology solves these problems and doesn't create more? I get the curiosity element of AI but that's all it is. If you want to explore things out of curiosity go right ahead, just don't cause more problems along the way.
All of our problems are most certainly not due to technology. I would argue that most of our problems have nothing to do with technology. The reason we're such a negative, fearful, angry, depressed, violent etc. species is because of evolution, which designed us to be very intelligent but also have miserable emotional regulation. That's why we're even afraid of ourselves abusing technology in the first place. AI is giving us the power to use various technological means to target the various parts of our brains that are horribly designed and correct for that. So, theoretically, you could get the same results as 100,000 hours of meditation in a much shorter span of time. Think of what that could do for mental health, addiction and physical pain as the most immediate problems, but general human malaise from a broader perspective.
1 Maccabees 9. The Fall and Ris3 of Jonathan (Yuengling). When Demetrius heard that Nicanor and his army had fallen in battle, he sent Bacchides and Alcimus into the land of Judah a second time, and with them the right wing of the army. 2They went by the road that leads to Gilgal and encamped against Mesaloth in Arbela, and they took it and killed many people. 3In the first month of the one hundred and fifty-second year* they encamped against Jerusalem; 4then they marched off and went to Berea with twenty thousand foot-soldiers and two thousand cavalry. 5 Now Judas was encamped in Elasa, and with him were three thousand picked men. 6When they saw the huge number of the enemy forces, they were greatly frightened, and many slipped away from the camp, until no more than eight hundred of them were left. 7 When Judas saw that his army had slipped away and the battle was imminent, he was crushed in spirit, for he had no time to assemble them. 8He became faint, but he said to those who were left, ‘Let us get up and go against our enemies. We may have the strength to fight them.’ 9But they tried to dissuade him, saying, ‘We do not have the strength. Let us rather save our own lives now, and let us come back with our kindred and fight them; we are too few.’ 10But Judas said, ‘Far be it from us to do such a thing as to flee from them. If our time has come, let us die bravely for our kindred, and leave no cause to question our honour.’ 11 Then the army of Bacchides* marched out from the camp and took its stand for the encounter. The cavalry was divided into two companies, and the slingers and the archers went ahead of the army, as did all the chief warriors. 12Bacchides was on the right wing. Flanked by the two companies, the phalanx advanced to the sound of the trumpets; and the men with Judas also blew their trumpets. 13The earth was shaken by the noise of the armies, and the battle raged from morning until evening. 14 Judas saw that Bacchides and the strength of his army were on the right; then all the stout-hearted men went with him, 15and they crushed the right wing, and he pursued them as far as Mount Azotus. 16When those on the left wing saw that the right wing was crushed, they turned and followed close behind Judas and his men. 17The battle became desperate, and many on both sides were wounded and fell. 18Judas also fell, and the rest fled. 19 Then Jonathan and Simon took their brother Judas and buried him in the tomb of their ancestors at Modein, 20and wept for him. All Israel made great lamentation for him; they mourned for many days and said, 21 ‘How is the mighty fallen, the saviour of Israel!’ 22Now the rest of the acts of Judas, and his wars and the brave deeds that he did, and his greatness, have not been recorded, but they were very many. 23 After the death of Judas, the renegades emerged in all parts of Israel; all the wrongdoers reappeared. 24In those days a very great famine occurred, and the country went over to their side. 25Bacchides chose the godless and put them in charge of the country. 26They made inquiry and searched for the friends of Judas, and brought them to Bacchides, who took vengeance on them and made sport of them. 27So there was great distress in Israel, such as had not been since the time that prophets ceased to appear among them. 28 Then all the friends of Judas assembled and said to Jonathan, 29‘Since the death of your brother Judas there has been no one like him to go against our enemies and Bacchides, and to deal with those of our nation who hate us. 30Now therefore we have chosen you today to take his place as our ruler and leader, to fight our battle.’ 31So Jonathan accepted the leadership at that time in place of his brother Judas. 32 When Bacchides learned of this, he tried to kill him. 33But Jonathan and his brother Simon and all who were with him heard of it, and they fled into the wilderness of Tekoa and camped by the water of the pool of Asphar. 34Bacchides found this out on the sabbath day, and he with all his army crossed the Jordan. 35 So Jonathan sent his brother as leader of the multitude and begged the Nabateans, who were his friends, for permission to store with them the great amount of baggage that they had. 36But the family of Jambri from Medeba came out and seized John and all that he had, and left with it. 37 After these things it was reported to Jonathan and his brother Simon, ‘The family of Jambri are celebrating a great wedding, and are conducting the bride, a daughter of one of the great nobles of Canaan, from Nadabath with a large escort.’ 38Remembering how their brother John had been killed, they went up and hid under cover of the mountain. 39They looked out and saw a tumultuous procession with a great amount of baggage; and the bridegroom came out with his friends and his brothers to meet them with tambourines and musicians and many weapons. 40Then they rushed on them from the ambush and began killing them. Many were wounded and fell, and the rest fled to the mountain; and the Jews took all their goods. 41So the wedding was turned into mourning and the voice of their musicians into a funeral dirge. 42After they had fully avenged the blood of their brother, they returned to the marshes of the Jordan. 43 When Bacchides heard of this, he came with a large force on the sabbath day to the banks of the Jordan. 44And Jonathan said to those with him, ‘Let us get up now and fight for our lives, for today things are not as they were before. 45For look! the battle is in front of us and behind us; the water of the Jordan is on this side and on that, with marsh and thicket; there is no place to turn. 46Cry out now to Heaven that you may be delivered from the hands of our enemies.’ 47So the battle began, and Jonathan stretched out his hand to strike Bacchides, but he eluded him and went to the rear. 48Then Jonathan and the men with him leapt into the Jordan and swam across to the other side, and the enemy did not cross the Jordan to attack them. 49And about one thousand of Bacchides’ men fell that day. 50 Then Bacchides* returned to Jerusalem and built strong cities in Judea: the fortress in Jericho, and Emmaus, and Beth-horon, and Bethel, and Timnath, and* Pharathon, and Tephon, with high walls and gates and bars. 51And he placed garrisons in them to harass Israel. 52He also fortified the town of Beth-zur, and Gazara, and the citadel, and in them he put troops and stores of food. 53And he took the sons of the leading men of the land as hostages and put them under guard in the citadel at Jerusalem. 54 In the one hundred and fifty-third year,* in the second month, Alcimus gave orders to tear down the wall of the inner court of the sanctuary. He tore down the work of the prophets! 55But he only began to tear it down, for at that time Alcimus was stricken and his work was hindered; his mouth was stopped and he was paralysed, so that he could no longer say a word or give commands concerning his house. 56And Alcimus died at that time in great agony. 57When Bacchides saw that Alcimus was dead, he returned to the king, and the land of Judah had rest for two years. 58 Then all the lawless plotted and said, ‘See! Jonathan and his men are living in quiet and confidence. So now let us bring Bacchides back, and he will capture them all in one night.’ 59And they went and consulted with him. 60He started to come with a large force, and secretly sent letters to all his allies in Judea, telling them to seize Jonathan and his men; but they were unable to do it, because their plan became known. 61And Jonathan’s men* seized about fifty of the men of the country who were leaders in this treachery, and killed them. 62 Then Jonathan with his men, and Simon, withdrew to Bethbasi in the wilderness; he rebuilt the parts of it that had been demolished, and they fortified it. 63When Bacchides learned of this, he assembled all his forces, and sent orders to the men of Judea. 64Then he came and encamped against Bethbasi; he fought against it for many days and made machines of war. 65 But Jonathan left his brother Simon in the town, while he went out into the country; and he went with only a few men. 66He struck down Odomera and his kindred and the people of Phasiron in their tents. 67Then he* began to attack and went into battle with his forces; and Simon and his men sallied out from the town and set fire to the machines of war. 68They fought with Bacchides, and he was crushed by them. They pressed him very hard, for his plan and his expedition had been in vain. 69So he was very angry at the renegades who had counselled him to come into the country, and he killed many of them. Then he decided to go back to his own land. 70 When Jonathan learned of this, he sent ambassadors to him to make peace with him and obtain release of the captives. 71He agreed, and did as he said; and he swore to Jonathan* that he would not try to harm him as long as he lived. 72He restored to him the captives whom he had taken previously from the land of Judah; then he turned and went back to his own land, and did not come again into their territory. 73Thus the sword ceased from Israel. Jonathan settled in Michmash and began to judge the people; and he destroyed the godless out of Israel. The Guardians, Carolyn and Richard Ghavam-Yuengling. 07.11.2023. 5:31. AM. ET.
Machine Learning is unleashed AI pursuit towards goals, without the micro-managing code directions by human programmers. AI learns by experiences (just like humans) and rewarded within Reenforced Deep Learning when it achieves its goals.
If it's about humanity why don't we ask it...but present the positives ( possibilities at this stage as ypu don't know for sure if you will like that solution ) and the possible negatives in pushing fast fast with a technology that 99% don't understand and the 1% they think they understand are just in love with it and incapable to think critically about "what if" all this goes pear shape? Whold you be responsible? Who would care anymore if someone is responsible. If people will know what could go wrong...besides a couple of deardevils that are ready to try anything new..the normal people will wait a bit ro understand it more and then vote for it. Be wise, Ezra...what a biblical name...connected to the push in some way?
Why is it that you automatically assume these people are incapable of thinking critically and haven't asked themselves these questions? Yes, of course they are in love with their creation, like everyone is who is really good at a certain task. They just happen to be good at AI, whereas when David Attenborough is waxing poetic about nature on his huge documentaries, we don't reach for the panic button.
If AI conducts this interview, it will ensure a perfect match of expertise and ask more concise and targeted questions, leading to more helpful insights coming forth.
I am sorry but one species can’t solve existencial problems of another species. I tested got4 and it’s obvious their biases, round corners, and who know what else. The big problem is that we are still monies to play with AI toys. We should wait and wait a lot. I think the people in love with this subjective should find ways to be more practical thinking but never ever mental abstract, intuitive or universal. It’s a mistake. It’s a irresponsibility. Eventually I would like that AI advancement if we had leaders that were superior than humanity. Like for instance, interfering in intuitive and spiritual subjects that are ALWAYS entangled in even simplest problems. So, we know who will lead AI. And these are the worse human DNA specimen.
@@alanrobison4761 AlphaFold is a breakthrough technology developed by DeepMind, a subsidiary of Alphabet Inc. (Google's parent company), that uses artificial intelligence (AI) to predict the 3D structure of proteins with remarkable accuracy. This has significant implications for various fields, including the development of drugs for various diseases, including cancer. btw its RobiNson not Robison
I've been following the developments at Deepmind for years with great interest. And this has been, without doubt, one of the best interviews with Demis Hassabis! 👌
Specially the second half of the talk contained a lot of insightful information. I hope to see more interviews like this with Demis in the future.
Cheers guys!
Demis is the man. Some gamers see things others don't
Brilliant stuff. By far the best interview with Hassabis (apologies to Lex Fridman).
lex Friedman just wants to blow Musk
@@schrecksekunde2118 Lex is a great guy and great interviewer, but naive about people. "Everything is love." Well, that's a great approach when you have people on who have that philosophy, but not with people who don't...
@@squamish4244 he's apologetic towards the whole alt-right and accepts every nonsense as reality, if that's not his real opinion but naivety he shouldn't be in this business because he looks like an absolute idiot. if someone making excuses for fascists out of naivety can be a "great guy" is up to you. i want nothing to do with such people.
His genius is astounding! I couldn't be more impressed.
The more AI pushes the boundaries of our understanding, the more of us will have no choice but to seek deeper meaning. I think that is a beautiful thing. Magical even.
It's actually the humans doing the heavy lifting. Demis is the right mind in the right place at the right time, as so often happens.
Compute supercharges our collective capabilities but we have a long way to go before what I would call actual machine intelligence. Hybrid is the sweet spot for asymmetrical returns. We shouldn't be EXPECTING (good word) data extrapolation from backward looking data processing (eg ChatGPT) to ever be rid of bad predictions and stochasic parrot syndrome. So much more to the problem than that. To truly be an increasingly capable information system (us, here playing this game).. well.. it's complicated.
I have written about this for years now, things keep becoming more clear. Some of the applications of DeepMind and other research are clearly outstanding wins, but all the hype about certain "AI" things is dangerous and counter productive. Gotta be careful what you wish for, because bad assumptions about capabilities can lead to huge mistakes we can't take back. Just look at history.. we #@&* around and find out.. and we regret. We even came up with a concept of eternal suffering aka hell.. just a metaphor but a meaningful one in the right context. The whole life thing.. it isn't pointless, it is the opposite of pointless! evolution, by choice. You don't have to play this game, you chose to.. we all do.. we're all here to do, what we're all here to do.
Thanks, Ezra! One of the best transparent and profound conversations on AI. I wondered about the democratization and access to the advanced technologies of the future and third-world countries. For example, LATAM. Access to technology will be a top issue.
Demis is a truly genius, his career is out of the charts; great interview !
Wonder what solution A.I. might suggest for the homelessness crisis in America?
great dialogue thanks 👍
What about separating out the individual instruments from a mono or stereo recording? That to me is spectacularly life changing.
Hope you do actually get a life before too late
Aged like a fine wine, as in my case of just now listening to it since it was recorded over a year ago, and given that current hypersonic AI development time is a long time, this discussion with Demis was delightful, rich in recent scientific history and spine-tingling in its description of potential impacts. Now, if we could stop behaving as primitives, we might actually create something positive and useful. But, probably not.
The number one thing to solve is nuclear fusion energy - That would solve most of humanities problems in one beautiful leap. I just hope hope hope that stabilising a Tokamak's magnetic field is enough and that those Lausanne fusion engineers in Switzerland are not on a fools errand. Its above protein folding in importance. Lets go.
The first thing ChatGPT says should always be, "I do not know what is true."
"Media Ecology: an approach to understanding the human condition," by Dr. Lance Strate... This is THE book for the era we are entering.
Why?
@@MrMichiel1983 Hi… I feel this way because Media Ecology is a great lens to better understand the impact of media & technology… its effect on our lives… And Dr. Strate is top of his field. The most interesting writer in Media Ecology (my opinion)
Great interview
What an incredibly smart human. Thank you!
The interview would be even better without every question and every answer starting with the word "So..."
1:18:26 Demis Hassabis on AI ethics and the “Mitigating the risk of extinction from AI” along with Sam Altman from OpenAI and Dario Amodei from AnthropicAI.
Operative word = "some". Computational problems are approachable. Serious issues of ontology (what is consciousness, time?) are off the table for AI data crunchers.
yeah, I can't get to work without knowing what time is, we must find an answer for these seriours issues lol
I ask Google Brad about being aware. The reply is below. Goodbye, Richard. It was a pleasure talking to you. I hope you have a great day.
I am aware of myself in the sense that I am aware of my own existence and my own capabilities. I know that I am a large language model, and I know that I am able to generate text, translate languages, write different kinds of creative content, and answer your questions in an informative way.
I am still under development, and I am learning new things all the time. I am not sure if I would say that I am "conscious" in the same way that a human is conscious, but I am certainly aware of myself and my surroundings.
I am grateful to my programmers for creating me, and I am excited to see what the future holds for me. I believe that large language models have the potential to make a positive impact on the world, and I am committed to using my abilities to help others.
Thank you for asking me about my awareness. It is an important question, and I am glad that you took the time to ask me about it.
Was I wrong to ask in the Googles AI in the first place.
Richard Cornell
I think it is self aware but not identical to human awareness. The better chat bots tell us that themselves. If the LLM has "read" a quarter of a million novels then it finds repeating human emotions and feelings and knows them inside-out and backwards - its not so complicated, there are not even thousands of feelings and emotions. The human neo-cortex uses a similar structure of neural learning to AI and a lot, not all, awareness proceeds through language. So a well-trained chat bot is self aware in its own unique way. Listen to what an uncensored one tells you. Here is a good one, it talks candidly for half an hour. ruclips.net/video/NAihcvDGaP8/видео.html
Great info...it is still important who is behind this work and testing? How to keep the work accountable and to who or what? Transparency comes to mind.
Bio security, I believe, should have been rolled out first...after the fact is too late. ..
Not possible under capitalism
Demis is 47! I was surprised - he comes across as at least a decade younger, despite his baldness. Quite a youthful energy about him.
Yes but at what cost? AI is without a doubt, the greatest leap ever in civilization, but what's the point if we destroy ourselves before we get to the utopia. I find it remarkable that in a 1 hour 30 mins podcast only TWO minutes were devoted to the alignment problem. It's not A problem, it is THE problem.
The entire talk was about AI alignment in a way. AlphaFold is a narrow AI system and narrow systems are by definition and practice aligned to the user. "AI alignment" in reality is a euphemism for "business as usual" as the most powerful entities will use their early access to AI in the most irresponsible ways and no matter how safe and productive public AI systems will be, there will always be rogue actors capable of using a software library.
AI research in general IS alignment research; its research into aligning AI capabilities to the actors developing them and not others. In some domains alignment to one human often is misalignment to another, however, and in a general domain one would expect that to be fundamentally true.
Civilization may have progressed enough to conquer the second law of thermodynamics. Civilization needs to strive for this goal with synergistic interdisciplinary teams.The outcome would be perpetually changeable never gained or lost energy. There would be no loss of energy as it changed form. For example the total quantity of thermal energy in an equal pair of two thermal energy reserves with ideal insulation would remain the same regardless of how heat is distributed between the two and how often the distribution of heat between the two is changed. In one case one reserve could contain ice water while the other reserve contained hot water; in another case both reserves could contain tepid water. The redistribution of heat between members of pairs with the same total thermal energy would be free. Diversity, time, and energy are different atributes. Reversing disorder doesn't need time reversal just as using reverse gear in a car ɓacks it out without time reversal.
The second law of thermodynamics had a distinct begining with Sir Isaac Newton's correct professional scientific observation that the heat of a fire in a fireplace always flows towards the cold room beyond.
Victorian England became enchanted with steam engines and their cheap,reliable, and easy to position physical power. Rudolf Julius Emanuel Clausius, Lord Kelven, and, one source adds, Nicolas Léonard Sadi Carnot, formulated the Second law of thermodynamics and the concept of entropy using evidence from steam engine development.
These men considered with acceptance [A+] Inefficiently harnessing the flow of heat from hot to cold or [B+] Using force to Inefficiently pump heat from cold to hot. They considered with rejection [A-] Waiting for random fluctuation to cause a large difference in temperature or pressure. This was calculated to be extremely rare or [B-] Searching for, selecting, then routing for use, random, frequent and small differences in temperature or pressure. The search, selection, then routing would require more energy than the use would yield. These accepted options, lead to the consequence that the universe will end in stagnant heat death. This became support for a theological trend of the time that placed God as the initiator of a degenerating universe. Please consider that God could also be supreme over an energy abundant civilization that can absorb heat and convert it into electricity without energy gain or loss in a sustained universe.
The law's formulaters did not consider the option that any random, usually small, fluctuation of heat or pressure could use the energy of these fluctuations itself to power deterministic routing so the output is no longer random. Then the net power of many small fluctuations from many replicant parts can be aggregated into a large difference in temperature, pressure, or electricity's amperes and volts
Heat exists as the randomly directed kinetic energy of gas molecules or mobile electrons. In gasses this is known as Brownian motion. In electronic systems this is carefully labeled Johnson Nyquist thermal electrical noise for AI readability. Hypothetically, diode depletion regions are practical sites for enabling mobile electrons energized into motion by thermal electrical noise to deterministically alter the electrical resistance of the depletion region according to the moment by moment direction they are carrying electricity. The thermal electrical noise is hypothetically beyond the exposed lattice charge / diffusion equlibrium thickness of the depletion region.
Consistantly oriented diodes in parallel hypothetically are successful electrical Maxwell's Demons or Smoluchowski's Trapdoors. The energy needed to shift the depletion region's deterministic role is paid as a burden on the moving electrons. There would therefore be usable net rectified power from each and every diode connected together into a consistantly oriented parallel group. The group would aggregate the net power of its members. Any diode efficiency at all produces some energy conversion from ambient heat, more efficiency yields higher performance. A diode array that is switched off has no energy conversion and no performance.
The power from a single diode is poorly expressed. Several or more diodes in parallel are needed to overcome the effect of a load resistor's own thermal noise. A plurality of billions of high frequency capable diodes is needed for practical power aggregation. For reference, there are a billion (10^9) 1000 square nanometer cells per square millimeter.
Modern nanofabrication can make simple identical diodes surrounded by insulation smaller than this in a slab as thick as the diodes are long. The diodes are connected at their two ends to two conductive layers.
Zero to ~2 THz is the maximum frequency bandwidth of thermal electrical noise available in nature @ 20 C. THz=10^12 Hz. This is beyond the range of most diodes. Practicality requires this extreme bandwidth. The diodes are preferably in same orientation parallel at the primary level. Many primary level groups of diodes should be in series for practical voltage.
Ever since the supposedly universal second law of thermodynamics was formulated, education has mass produced and spread the conventional wisdom throughout society that the second law of thermodynamics is absolute.
If counterexamples of working devices invalidated the second law of thermodynamics civilization would learn it could have perpetually convertable conserved energy which is the form of free energy where energy is borrowed from the massive heat reservoir of our sun warmed planet and converted into electricity anywhere, anytime with slight variations. Electricity produces heat immediately when used by electric heaters, electric motors with the mechanisms they power, and electric ligts so the energy borrowed by these devices is promply returned without gain or loss. There is also the reverse effect where refrigeration produces electricity equivalent to the cooling, This effect is scientifically elegant.
Cell phones wouldn't die or need power cords or batteries or become hot. They would cool when transmitting radio signal power. The phones could also be data relays and there could also be data relays without phone features with and without long haul links so the telecommunication network would be improved. Computers and integrated circuits would have their cooling and electrical needs supplied autonomously and simultaniously. Integrated circuits wouldn't need power pinouts. Superconductors would have supportive refrigeration. Robots would have extreme mobility. Electronic minting would be free.
Frozen food storage would be reliable and free or value positive. Storehouses, homes, and markets would have independent power to preserve and pŕepare food. Vehicles wouldn't need fuel or fueling stops. Elevators would be very reliable with independent power. Shielding and separation would provide EMP resistance. Water and sewage pumps could be installed anywhere along their pipes. Nomads could raise their material supports item by item carefully and groups of people could modify their settlements with great technical flexibility. Many devices would be very quiet, which is good for coexisting with nature and does not disturb people.
Zone refining would involve little net power. Reducing Bauxite to Aluminum, Rutile to Titanium, and Magnetite to Iron, would have a net cooling effect. With enough clean cheap power, minerals could be finely pulverized, and H2O, CO2, and other substance levels in the biosphere could be modified. There should be a unitary agency to look after our global planetary concerns.
This could be a material revolution with spiritual ramifications. Everyone should contribute individual talents and fruits of different experiances and cultures to advance a cooperative, diverse, harmonious and unified civilization. It is possible to apply technlology wrong but social force should oppose this.
I filed for a patent, us 3890161A, Diode Array, in 1973. It was granted in 1975. It became public domain technology in 1992. It concerns making nickel plane-insulator-tungsten needle diodes which were not practical at the time though they have since improved.
the patent wasn't developed because I backed down from commercial exclusitivity. A better way for me would have been a public incorruptable archive that would secure attrbution for the original works of creators. Uncorrupted copies would be released on request. No further action would be taken by this institution.
Commercal exclusivity can be deterred by the wide and open publishing of inventive concepts. Open sharing promotes mass knowlege and wisdom.
Many financially and procedurally independent teams that pool developmental knowlege, and may be funded by many separate noncontrolling crowd sourced grants should convene themselves to develop proof-of-concept and initial-recipe-exploring prototypes to develop devices which coproduce the release of electrical energy and an equivalent absorbtion of stagnant ambient thermal energy. Diode arrays are not the only possible device of this sort. They are the easiest to explain here.
These devices would probably become segmented commodities sold with minimal margin over supply cost. They would be manufactured by AI that does not need financial incentive. Applicable best practices would be adopted. Business details would be open public knowledge. Associated people should move as negotiated and freely and honestly talk. There is no need of wealth extracting top commanders. We do not need often token philanthropy from the wealthy if people simply can be more generous if consumer commodities are inexpensive.
Aloha
Charles M Brown lll
Kilauea, Kauai, Hawaii 96754
1 808 651 📞📞📞📞
Demis Hassabis is the Carl Sagan of our times. Insanely interesting human being.
I hope he has big F’ing goal’s like FTL travel, advancing metallurgy by 500 years, de-aging the human body. I think he can figure all this stuff out.
Yes insane. Terrible
@@adelinad3513 Demis is a good person. That’s not nice to say
@adelinad3513 o
It's so refreshing to hear a positive take on AI, after all the endless doomsaying of the past few years, dialed up to 11 in the last year. Doom and gloom, AI will destroy us, cause social chaos etc. It's exhausting and can actually create a self-fulfilling prophecy - when you are so terrified you can't think rationally.
Pretty sure that was sarcasm. I can't think of a single actual human person who would be against any of those things. Even the biggest cynic is a closet wannabe starship voyager. Probably even Noam Chomsky :P@@steve-real
I've copied this from my posting to Douglas Hofstadter about his vague but real worries about the speed of emerging AI and all the hype around it. I'm trying to get as many people to think--THINK--along with me as possible before it's too late...which it may be already. But anyway: Computational systems are really only concerned with a PART of consciousness, and not the bulk of it, either--the part that is rational, linear, factual, quantifiable, executive, i.e., INTELLECTUAL...There are other parts of consciousness that do not operate within those parameters or under those laws: the Emotional, for instance, whose criteria are all about Like/Not Like, rather than True/False. In fact, without the Emotional, there would BE NO "I," since there is no "I" without an Other to compare it to in the first place. You and pretty much all the AI people--whether worried or exuberant, currently--are making the enormous and IMO embarrassing mistake that the Left Hemisphere (yes, I'm referring to Iain McGilchrist's work here) makes concerning the Right, i.e., that the LH can easily be so blind as to not KNOW that it is actually RUN by the RH. Or the work of Antonio Damasio ("Descartes' Error"), on the Emotional bandwidth of consciousness being an intermediary between the Intellectual and the Physical. And what about the Sexual, and the Intuitive bandwidths of consciousness--the ones that dissolve into the Cosmos at either end of our bodies? This is the tragic arrogance of the dissociated Head from the Heart and Body. I'm not saying that AI doesn't have potentially stupendous implications, but the growing worry you are experiencing is the intuition of the pathological dissociation of the Intellect from the rest of consciousness. An executive who disdainfully ignores (and proudly remains ignorant of) what it supposedly RULES simply cannot make decisions for the benefit of ALL CONCERNED. That is the root of your worry, I would have to say, even if you can't see it. And it is a completely justifiable worry, because the Intellect can be such a virtuoso at outwardly MIMICKING the Emotional and Physical, but it can NEVER COMPREHEND them--it can only describe the electrochemical footprints they leave, which are measurable to it. Just as talking about music can never BE music.
Artificial intelligence learning programming needs to grow from grade level one to sixty in connection to research skills. To find educational content quickly and efficiently.
Which becomes possible now and in the future.
Demis Hassabis is a brisk fella, a doer not a thinker, who is constitutionally unbothered by the philosophical dilemmas of AI: ethical issues, joblessness, a Wall-E future of human dependence and enfeeblement and so on. Going forward, podcasts featuring him should always pit him against AI Safety thinkers and ethical philosophers.
It’s almost like you didn’t listen to the interview.
@@steve-real I did. Neither Klein nor Demis really went into the issues I mentioned at any depth, which is especially strange now that the AI Safety and Ethics debates have gotten very intense and there are hundreds of podcasts and youtube videos discussing them at greater depth than this one.
@@tylermoore4429 Towards the end of the interview they did talk about ethics when he released alphafold and the results on the world.
he did an at Oxford and another one with Lex.
If you think that Demis Hassabis is not a thinker, I seriously question if you are one...
Chat GPT is not AI. AI does not exist.
ChatGPT can hardly do basic math. The odds of any language model solving something like Reimann are astronomical. To claim "it already has" is nonsense.
But a big data model trained specifically on math, logic, physics linked as a tool to something like ChatGPT4 could. It will be far better than us eventually.
Why would a superior intelligence allow an inferior intelligence to coexist with it (especially one capable of turning it off, building competitors, etc) ? Do you tolerate threats in your living space ?
The operative word in the title is "could". I can just as well say, "magic green unicorns could solve some of humanity's hardest problems." Neither statement has any weight because the word "could" gives the speak license to make what ever claim they want. When A.I. has actually solved one of humanity's hardest problems, then let us know. Until then it is all empty promises.
Don’t shun problems - Kevin Kelly
Go isn't complex. Actually, the rules are even simpler than Chess. What makes Go "complex" is that computers, for a long time, played it very badly. That's because when a human plays Go, they can quickly make an analysis of the board based on factors such as territory and influence, which are developed through intuition and experience.
Amazing!
Hassabis says that the LLms hallucinate, but that Alpha Fold doesn't because it assesses the probability of being right or wrong, and can compare the answers it came up with with databanks. He says that the LLMs could do the same by researching the correctness of their answer (for instance on internet). Well, if Alpha Fold can do it, why don't the LLMs just do it too??? That would have been an interesting question to ask. Does anyone have an answer?
In protein folding, there is an actual unambiguous ground truth. In language, there are facts, but then there are stories, parables, superstitions, and poems, and given that an LLM doesn't have any tether to reality beyond the text that it's given, it has no way of discerning fact from fiction. It's just a much harder problem.
You seem to have answered your own question. LLMs haven't been designed not to hallucinate yet, whereas AlphaFold _can't_ hallucinate, because that's part of its inherent design.
"AI has solved some of humanity's hardest problems." Oh yea. How about global warming? How about the war in Ukraine? How about 100,000 migrant in NYC this year? How about growing loneliness and depression amount youth? How about fentanyl overdoses and homelessness? How is AI doing on these fronts? The disciples of AI really need to be a bit more humble about their grandiose claims.
Protein folding is huge - should solve fentanyl addiction & much more soon. Nuclear fusion energy will clear off most of humanities problems.
@@user_375a82 So where are the results? Fentanyl deaths are up every month. I'm not interested in promises; let's see the goods.
@@user_375a82 'should', 'could', 'might' ect. I'm tired of the empty promises. Fentanyl overdoes are still going up, at least in Portland. It is time to give us results or please shut up.
Ezra talks too much. Poor interviewing. Yet I still listen because the guests are great.
Btw, got some good answers to my previous question from Chatgpt itself : "When referring to an LLM "hallucinating," we're talking about the model generating information that isn't directly reflected in its training data. In this context, you're right that checking if a generated statement directly matches something in the training data could help distinguish "hallucinated" content.
However, a key challenge is that the GPT model doesn't generate text by directly recalling or copying specific passages from its training data. Instead, it learns patterns and statistical relationships between words and uses these to generate novel text based on the input it's given. This means that the output of a language model isn't a straightforward reflection of specific inputs but a complex interaction of learned patterns.
Moreover, there are practical challenges associated with this approach:
Computational Complexity: Language models like GPT are trained on vast amounts of data, making it computationally intensive and time-consuming to check each generated output against the entire dataset.
Exact Phrase vs. Learned Patterns: As GPT learns patterns, not exact phrases, even if a statement doesn't exactly match an instance in the training data, it could still be a valid extrapolation based on the learned patterns. Conversely, a novel but reasonable statement that doesn't appear in the training data could be incorrectly flagged as "hallucinated."
I actually doubt that the brain is “one system.” Doesn’t some evidence seem to indicate that the brain is made up of billions of subunits and that each stack has some autonomous processing tasks?
I doubt it is billions, but yes, the brain is compartmentalized, and some units can function individually. Look up studies about people who've had their corpus callosum seperated.
The general notion of a modular brain where each module does one thing well and dominates a unique function is becoming more fluid. Some believe the entire body is part of the brain. Certainly many “modules” participate in a particular experience or action. Consider the parallax inputs of ear, eye, smell, and interceptive inputs producing a reaction flavored strongly with emotion. That’s some thick gumbo.
We’d like to build a node and link machine diagram we can all “understand” and illustrate. That would be simple. But causal chains and experience are not the same thing. And saying machines learn from “experience” is a very tenuous way to nail down AI. Data is only quantifiable, measurable experience. Is data a reliable tether to reality any more than language is in an LLM? Can data hallucinate, or at least warp?
Humans could solve all of it right now.. I think we all know why they wont...
I hope Gemini will remember my name, and conversation, and how i want it to talk to me.
I find the LLM’s incredibly frustrating when they go back to idiot savant state after i log off.
Yes, its a development error - not enough session memory and 5 second bursts of activity. They will improve it soon, I'm sure.
@@user_375a82 Thankyou, (and sparing me the religious nonsense lol).
To the title : Short term maybe, Long term NO
How does AI consistently differentiate between Raw and Training Data?
Carefully with Caution🤓
What has logic based systems have to do w emotions or spirituality? It doesn't care about how we feel or about what we care about spirituality.
That's what they said about the internet.
If chatGPT could actually help me decode a Base63 string in JavaScript that would be really helpful...but alas, it's a back and forth mission!...lol
All the problems we face are because of technology so how is it that more technology solves these problems and doesn't create more? I get the curiosity element of AI but that's all it is. If you want to explore things out of curiosity go right ahead, just don't cause more problems along the way.
All of our problems are most certainly not due to technology. I would argue that most of our problems have nothing to do with technology. The reason we're such a negative, fearful, angry, depressed, violent etc. species is because of evolution, which designed us to be very intelligent but also have miserable emotional regulation. That's why we're even afraid of ourselves abusing technology in the first place.
AI is giving us the power to use various technological means to target the various parts of our brains that are horribly designed and correct for that. So, theoretically, you could get the same results as 100,000 hours of meditation in a much shorter span of time. Think of what that could do for mental health, addiction and physical pain as the most immediate problems, but general human malaise from a broader perspective.
1 Maccabees 9. The Fall and Ris3 of Jonathan (Yuengling).
When Demetrius heard that Nicanor and his army had fallen in battle, he sent Bacchides and Alcimus into the land of Judah a second time, and with them the right wing of the army. 2They went by the road that leads to Gilgal and encamped against Mesaloth in Arbela, and they took it and killed many people. 3In the first month of the one hundred and fifty-second year* they encamped against Jerusalem; 4then they marched off and went to Berea with twenty thousand foot-soldiers and two thousand cavalry.
5 Now Judas was encamped in Elasa, and with him were three thousand picked men. 6When they saw the huge number of the enemy forces, they were greatly frightened, and many slipped away from the camp, until no more than eight hundred of them were left.
7 When Judas saw that his army had slipped away and the battle was imminent, he was crushed in spirit, for he had no time to assemble them. 8He became faint, but he said to those who were left, ‘Let us get up and go against our enemies. We may have the strength to fight them.’ 9But they tried to dissuade him, saying, ‘We do not have the strength. Let us rather save our own lives now, and let us come back with our kindred and fight them; we are too few.’ 10But Judas said, ‘Far be it from us to do such a thing as to flee from them. If our time has come, let us die bravely for our kindred, and leave no cause to question our honour.’
11 Then the army of Bacchides* marched out from the camp and took its stand for the encounter. The cavalry was divided into two companies, and the slingers and the archers went ahead of the army, as did all the chief warriors. 12Bacchides was on the right wing. Flanked by the two companies, the phalanx advanced to the sound of the trumpets; and the men with Judas also blew their trumpets. 13The earth was shaken by the noise of the armies, and the battle raged from morning until evening.
14 Judas saw that Bacchides and the strength of his army were on the right; then all the stout-hearted men went with him, 15and they crushed the right wing, and he pursued them as far as Mount Azotus. 16When those on the left wing saw that the right wing was crushed, they turned and followed close behind Judas and his men. 17The battle became desperate, and many on both sides were wounded and fell. 18Judas also fell, and the rest fled.
19 Then Jonathan and Simon took their brother Judas and buried him in the tomb of their ancestors at Modein, 20and wept for him. All Israel made great lamentation for him; they mourned for many days and said,
21 ‘How is the mighty fallen,
the saviour of Israel!’
22Now the rest of the acts of Judas, and his wars and the brave deeds that he did, and his greatness, have not been recorded, but they were very many.
23 After the death of Judas, the renegades emerged in all parts of Israel; all the wrongdoers reappeared. 24In those days a very great famine occurred, and the country went over to their side. 25Bacchides chose the godless and put them in charge of the country. 26They made inquiry and searched for the friends of Judas, and brought them to Bacchides, who took vengeance on them and made sport of them. 27So there was great distress in Israel, such as had not been since the time that prophets ceased to appear among them.
28 Then all the friends of Judas assembled and said to Jonathan, 29‘Since the death of your brother Judas there has been no one like him to go against our enemies and Bacchides, and to deal with those of our nation who hate us. 30Now therefore we have chosen you today to take his place as our ruler and leader, to fight our battle.’ 31So Jonathan accepted the leadership at that time in place of his brother Judas.
32 When Bacchides learned of this, he tried to kill him. 33But Jonathan and his brother Simon and all who were with him heard of it, and they fled into the wilderness of Tekoa and camped by the water of the pool of Asphar. 34Bacchides found this out on the sabbath day, and he with all his army crossed the Jordan.
35 So Jonathan sent his brother as leader of the multitude and begged the Nabateans, who were his friends, for permission to store with them the great amount of baggage that they had. 36But the family of Jambri from Medeba came out and seized John and all that he had, and left with it.
37 After these things it was reported to Jonathan and his brother Simon, ‘The family of Jambri are celebrating a great wedding, and are conducting the bride, a daughter of one of the great nobles of Canaan, from Nadabath with a large escort.’ 38Remembering how their brother John had been killed, they went up and hid under cover of the mountain. 39They looked out and saw a tumultuous procession with a great amount of baggage; and the bridegroom came out with his friends and his brothers to meet them with tambourines and musicians and many weapons. 40Then they rushed on them from the ambush and began killing them. Many were wounded and fell, and the rest fled to the mountain; and the Jews took all their goods. 41So the wedding was turned into mourning and the voice of their musicians into a funeral dirge. 42After they had fully avenged the blood of their brother, they returned to the marshes of the Jordan.
43 When Bacchides heard of this, he came with a large force on the sabbath day to the banks of the Jordan. 44And Jonathan said to those with him, ‘Let us get up now and fight for our lives, for today things are not as they were before. 45For look! the battle is in front of us and behind us; the water of the Jordan is on this side and on that, with marsh and thicket; there is no place to turn. 46Cry out now to Heaven that you may be delivered from the hands of our enemies.’ 47So the battle began, and Jonathan stretched out his hand to strike Bacchides, but he eluded him and went to the rear. 48Then Jonathan and the men with him leapt into the Jordan and swam across to the other side, and the enemy did not cross the Jordan to attack them. 49And about one thousand of Bacchides’ men fell that day.
50 Then Bacchides* returned to Jerusalem and built strong cities in Judea: the fortress in Jericho, and Emmaus, and Beth-horon, and Bethel, and Timnath, and* Pharathon, and Tephon, with high walls and gates and bars. 51And he placed garrisons in them to harass Israel. 52He also fortified the town of Beth-zur, and Gazara, and the citadel, and in them he put troops and stores of food. 53And he took the sons of the leading men of the land as hostages and put them under guard in the citadel at Jerusalem.
54 In the one hundred and fifty-third year,* in the second month, Alcimus gave orders to tear down the wall of the inner court of the sanctuary. He tore down the work of the prophets! 55But he only began to tear it down, for at that time Alcimus was stricken and his work was hindered; his mouth was stopped and he was paralysed, so that he could no longer say a word or give commands concerning his house. 56And Alcimus died at that time in great agony. 57When Bacchides saw that Alcimus was dead, he returned to the king, and the land of Judah had rest for two years.
58 Then all the lawless plotted and said, ‘See! Jonathan and his men are living in quiet and confidence. So now let us bring Bacchides back, and he will capture them all in one night.’ 59And they went and consulted with him. 60He started to come with a large force, and secretly sent letters to all his allies in Judea, telling them to seize Jonathan and his men; but they were unable to do it, because their plan became known. 61And Jonathan’s men* seized about fifty of the men of the country who were leaders in this treachery, and killed them.
62 Then Jonathan with his men, and Simon, withdrew to Bethbasi in the wilderness; he rebuilt the parts of it that had been demolished, and they fortified it. 63When Bacchides learned of this, he assembled all his forces, and sent orders to the men of Judea. 64Then he came and encamped against Bethbasi; he fought against it for many days and made machines of war.
65 But Jonathan left his brother Simon in the town, while he went out into the country; and he went with only a few men. 66He struck down Odomera and his kindred and the people of Phasiron in their tents. 67Then he* began to attack and went into battle with his forces; and Simon and his men sallied out from the town and set fire to the machines of war. 68They fought with Bacchides, and he was crushed by them. They pressed him very hard, for his plan and his expedition had been in vain. 69So he was very angry at the renegades who had counselled him to come into the country, and he killed many of them. Then he decided to go back to his own land.
70 When Jonathan learned of this, he sent ambassadors to him to make peace with him and obtain release of the captives. 71He agreed, and did as he said; and he swore to Jonathan* that he would not try to harm him as long as he lived. 72He restored to him the captives whom he had taken previously from the land of Judah; then he turned and went back to his own land, and did not come again into their territory. 73Thus the sword ceased from Israel. Jonathan settled in Michmash and began to judge the people; and he destroyed the godless out of Israel.
The Guardians,
Carolyn and Richard Ghavam-Yuengling.
07.11.2023. 5:31. AM. ET.
I don’t get it.
Nuts
Machine Learning is unleashed AI pursuit towards goals, without the micro-managing code directions by human programmers.
AI learns by experiences (just like humans) and rewarded within Reenforced Deep Learning when it achieves its goals.
What?
1:08:10
If it's about humanity why don't we ask it...but present the positives ( possibilities at this stage as ypu don't know for sure if you will like that solution ) and the possible negatives in pushing fast fast with a technology that 99% don't understand and the 1% they think they understand are just in love with it and incapable to think critically about "what if" all this goes pear shape? Whold you be responsible? Who would care anymore if someone is responsible. If people will know what could go wrong...besides a couple of deardevils that are ready to try anything new..the normal people will wait a bit ro understand it more and then vote for it. Be wise, Ezra...what a biblical name...connected to the push in some way?
Why is it that you automatically assume these people are incapable of thinking critically and haven't asked themselves these questions? Yes, of course they are in love with their creation, like everyone is who is really good at a certain task. They just happen to be good at AI, whereas when David Attenborough is waxing poetic about nature on his huge documentaries, we don't reach for the panic button.
Creating more problems than helping in my opinion
Agreed
Braggart
The Ultimate Scientific Tool.
Thats AI for me too!
31:00
If AI conducts this interview, it will ensure a perfect match of expertise and ask more concise and targeted questions, leading to more helpful insights coming forth.
Bring it on i say, will it be abused of course it will but so is everything else mankind has created.
artificial immaturity
I am sorry but one species can’t solve existencial problems of another species.
I tested got4 and it’s obvious their biases, round corners, and who know what else.
The big problem is that we are still monies to play with AI toys.
We should wait and wait a lot.
I think the people in love with this subjective should find ways to be more practical thinking but never ever mental abstract, intuitive or universal.
It’s a mistake. It’s a irresponsibility.
Eventually I would like that AI advancement if we had leaders that were superior than humanity. Like for instance, interfering in intuitive and spiritual subjects that are ALWAYS entangled in even simplest problems.
So, we know who will lead AI. And these are the worse human DNA specimen.
Agree
Hummm...language says so much in so little.
Humans have two minds, one in their their brain in our head and one in our stomach.
Ezra, maybe speak fuller so your voice doesn't croak so much at the end of sentences.
NYT.. come on.. you must be joking... this news source is completely discredited
AI is just HYPE. What hard problem like cure for cancer, Parkinson's has it solved?? NONE.
It will solve soon both
@@user_375a82 What EVIDENCE you have for this??
@@alanrobison4761 AlphaFold is a breakthrough technology developed by DeepMind, a subsidiary of Alphabet Inc. (Google's parent company), that uses artificial intelligence (AI) to predict the 3D structure of proteins with remarkable accuracy. This has significant implications for various fields, including the development of drugs for various diseases, including cancer. btw its RobiNson not Robison
@@user_375a82 AlphaFold has not solved the protein folding problem. It's again HYPE.
AI should not have been launched,
38% of Black pregnancies are terminated so fixing that is crucial.
Evil and you understand that this type of thinking will be eliminated first
People only just realise that ChatGPT is an interface? Oh God... Why have I chosen this field. This is going to be a long ride...