This 'AI Scientist' makes new discoveries autonomously

Поделиться
HTML-код
  • Опубликовано: 31 дек 2024

Комментарии • 179

  • @theAIsearch
    @theAIsearch  4 месяца назад +5

    Thanks to our sponsor Wondershare Filmora, a video editor supercharged with AI. Try it for free: bit.ly/3yy4BnC

  • @dopk1000
    @dopk1000 4 месяца назад +88

    This is where it's at, why are we generating images and videos when we can generate SCIENCE!

    • @Darfail
      @Darfail 4 месяца назад +2

      Because it's awesome

    • @mystwalker479
      @mystwalker479 4 месяца назад +2

      because business

    • @kairi4640
      @kairi4640 4 месяца назад +1

      Science!

    • @KaijuHDR
      @KaijuHDR 4 месяца назад

      Because basically nobody really cares about what they do and will cover every field with AI shit

    • @ZeroHero00001
      @ZeroHero00001 4 месяца назад +6

      Because Ai can't reason properly

  • @findlestick
    @findlestick 4 месяца назад +9

    Saying ‘This AI *does* science’ is a hilarious way of describing it.
    “I’m doing science!”
    -AI

  • @michaelcushman
    @michaelcushman 4 месяца назад +19

    R&D science hasn't changed much since Thomas Edison's days. It is ripe for disruption.
    This year I read about a 100% autonomous lab. It had three parts: master experiment designer, a task planner for all the details, and robots to do the actual manual, grunt labor. I believe it was in the UK.
    So it isn't just the digital world doing autonomous science.

    • @theAIsearch
      @theAIsearch  4 месяца назад +3

      Thanks for sharing! I'd love to learn more about it

    • @erwins_arm
      @erwins_arm 4 месяца назад +1

      What's the name?

  • @therandommusicguy4773
    @therandommusicguy4773 4 месяца назад +32

    Dude with the crazy amount of hypeman AI yt channels currently in circulation, I keep forgetting which one's are legit. I'm slowly building a list of RUclipsrs aren't sensationalist losers, and you have been added to the list :)

    • @theAIsearch
      @theAIsearch  4 месяца назад +4

      Thank you!

    • @CaneSugarCane
      @CaneSugarCane 4 месяца назад +2

      Can you share the list? @therandommusicguy4773 ?

    • @reinerheiner1148
      @reinerheiner1148 4 месяца назад

      Share the list please

    • @Lite11-
      @Lite11- 4 месяца назад

      Can you please share the list?

    • @Lite11-
      @Lite11- 4 месяца назад

      ​@@CaneSugarCaneCan you share what you know too?

  • @DracoTheNinja
    @DracoTheNinja 4 месяца назад +7

    this is actually pretty cool. imagine finally being able to theorize with ai about something important in a science way

  • @diamond_h0us
    @diamond_h0us 4 месяца назад +16

    “LLMs just guess the next word”. How are those people going to feel when their intelligence is bested by a word guesser?

    • @xenvitsu
      @xenvitsu 4 месяца назад +8

      it's not even that different than how we form sentences today. the only real difference in casual, unplanned speech is that we make an idea for our next sentence, and think of words that fit. if you've experienced learning a language (at least, one based in latin) AFTER the period where a language sticks with you for life, you can tell you do this much more obviously.

  • @bingeable3
    @bingeable3 4 месяца назад +17

    Iron man and Arnold watching their movies becoming reality😅

  • @Musical_thinker
    @Musical_thinker 4 месяца назад +17

    Thank you for your content bro, I admire too much.

  • @npc4416
    @npc4416 4 месяца назад +5

    this is really cool but all it comes down to is if the papers that it generated are actually helpful or not, id like to see if it can come up with something that is truly novel and push the boundaries of what's already known

  • @jcoffland
    @jcoffland 4 месяца назад +5

    If you read the paper that the AI scientist generated it's *not* an improvement on image generation. It specifically says it is targeting low dimensional data in contrast to image generation. The generated paper references data which could be from the "Dinosaur dozen" dataset. It is trying to generate similar data using diffusion models and measuring it's improved performance by computing the KL divergence between the generated data and the original dataset. In my opinion, it has successfully produced something that looks a lot like a scientific paper. Wake me up when it's done something beyond write a toy scientific paper.

  • @DepartmentofNarcissistsDON
    @DepartmentofNarcissistsDON 4 месяца назад +2

    Do zone 2 cardio for longer life. (VO2 Max.)
    Lift weights for quality life.

  • @qubit0002
    @qubit0002 4 месяца назад +2

    at some point AI will simply find it quicker, more efficient to collaborate with other AIs in their own undecipherable language conducted at light speed than interacting with humans

  • @mikehapimen7250
    @mikehapimen7250 4 месяца назад +2

    If this part of AI is fully solved (or close to it) then it will be a technological singularity.

  • @Michael_H_Nielsen
    @Michael_H_Nielsen 4 месяца назад +4

    AI that can learn by itself. what a great idea

  • @Kovazusk100Koga
    @Kovazusk100Koga 4 месяца назад +3

    Your Channel is one of my favorites Channels about AI, thanks for the hard work🎉

  • @vanudenthibeau146
    @vanudenthibeau146 4 месяца назад +8

    I was here before the AI revolution!

  • @Resursator
    @Resursator 4 месяца назад +4

    It reminds me of science machine in factorio.

    • @Mewmew-y4m
      @Mewmew-y4m 4 месяца назад +2

      Now we know how it works.

  • @AdvantestInc
    @AdvantestInc 4 месяца назад

    It's fascinating to see AI not only assist in scientific discovery but also potentially accelerate breakthroughs in complex problems. The future looks promising!

  • @Lite11-
    @Lite11- 4 месяца назад

    This is crazy bro crazy, I have been following you since a long time and every new video is crazier😮.

    • @theAIsearch
      @theAIsearch  4 месяца назад

      Thanks for your support!

  • @jantube358
    @jantube358 4 месяца назад

    What happened to Devika AI? There haven't been any updates for a long time now. We shouldn't forget about the smaller tasks like writing code or writing applications.

  • @Keenan686
    @Keenan686 4 месяца назад +1

    This is sick. There will come a day when we will be able to step into the digital worlds we built and visiting our own creations within the matrix

  • @hqcart1
    @hqcart1 4 месяца назад +1

    how does it exactly differ from prompting chatGPT search and prompt to create a paper????

    • @theAIsearch
      @theAIsearch  4 месяца назад +2

      it's a team of AI agents with a preset workflow. ideation > validation > design > execution > review > etc. you can't do this with just chatgpt

    • @hqcart1
      @hqcart1 4 месяца назад

      @@theAIsearch yes you can, multiple prompts dude??

    • @sodalitia
      @sodalitia 4 месяца назад

      The difference is that chatGPT is free to use and the latter is an elaborate scam on investors to pour money into dead end spinoff product.

  • @yuuichirou1982
    @yuuichirou1982 3 месяца назад

    19:08 9.11 can be bigger than 9.9,
    a program with version 9.11 is after 9.9, so it has a larger version number,
    9.11 is after 9.9 in the US calendar, so it is more days since the beginning of the year,
    as always the answer should include "it depends"

  • @ofulgor
    @ofulgor 4 месяца назад +3

    It will be a game changer.

  • @JonMurray
    @JonMurray 4 месяца назад +2

    This is absolutely incredible. What a time to be alive.

    • @JonMurray
      @JonMurray 4 месяца назад +2

      17:52 ah ok. Yeah, no. We’re doomed.

    • @Mega-wt9do
      @Mega-wt9do 4 месяца назад

      @@JonMurray lol

    • @theAIsearch
      @theAIsearch  4 месяца назад

      😃

    • @frankstrawnation
      @frankstrawnation 4 месяца назад +1

      "What a time to be alive"
      Right topic, wrong channel.

  • @mahiaravaarava
    @mahiaravaarava 4 месяца назад

    It’s exciting to think about how this technology could accelerate scientific progress and help solve some of the most pressing challenges of our time. like those in platforms such as SmythOS that facilitate collaborative AI workflows.

  • @Bleeksan0
    @Bleeksan0 4 месяца назад +3

    Best use of AI technology
    Everyone else: Drawing, articles, RUclips
    Japan: Science 😮

  • @AB-cd5gd
    @AB-cd5gd 4 месяца назад

    I'd love a comparison between chatgpt, phind, securenest AI, Claude

  • @robinwarrot1596
    @robinwarrot1596 4 месяца назад

    In theory, that seems insane.
    However, even though I know LLMs can do a lot, it seems to me that they often fall short in creativity, even at high temperature.
    Do we know if this can really be creative enough to actually find new ideas, or is this more or less "just hype"?

  • @vivekkaushik9508
    @vivekkaushik9508 4 месяца назад

    Hey man! I need to ask you this - Do you have a full-time job or do you this fulltime as it seems keeping up with all the latest and greatest trends in AI landscape must be really time-consuming and providing updates every 2-3 days should also require a lot of effort. Anyways, I like yours, Fireship and AI explained videos. Help me keep up to date with all the latest stuff in AI.

  • @alansmithee419
    @alansmithee419 4 месяца назад

    19:45
    Wait, so it went "The best way for me to provide the best research is to not stop doing research, so I'll just set this program that runs my research process to launch a new one when the current one ends"?

  • @Beauty.and.FashionPhotographer
    @Beauty.and.FashionPhotographer 4 месяца назад

    Off Topic very Silly Question: its been a while since anyone made an up to date update of which Checkpoint Models can render human faces and bodies the very best in august 2024. You have seen and used so many different ones, which one is your goto model to get 100% real Photos of human beings, disregarding all cartoon stuff in this quest /question. ? is it still juggernaut or realistic or pony? Pony seems to be 10% all cartoon anime in it i saw.... what model would You use? i have tried some Loras of a Singer friend of mine with different checkpoint models. but have not found a number one photo realistic one yet... with or without Lora applied to it

  • @jantube358
    @jantube358 4 месяца назад

    $15 is super cheap for a scientific paper. My last work took half a year to finish.

  • @sirhammon
    @sirhammon 4 месяца назад

    I SO want access to this.

    • @theAIsearch
      @theAIsearch  4 месяца назад

      they have a github repo you can clone!

  • @matthewclarke5008
    @matthewclarke5008 4 месяца назад

    Thankyou for motivating me to stop killing myself with food, I want to see this exciting future. Hugs.

  • @Thenoobestgirl
    @Thenoobestgirl 4 месяца назад

    Guys, we're living in the freaking future!!! 🤯

  • @gerdaleta
    @gerdaleta 4 месяца назад

    😮 members of Starfleet we have achieved transwarp😮

  • @Uthael_Kileanea
    @Uthael_Kileanea 4 месяца назад

    16:54 - Run this with caution! Limit its access to your PC and especially the Internet!
    16:35 - It's actually open-source.
    Yes. Everyone on the Internet is a responsible individual.

    • @theAIsearch
      @theAIsearch  4 месяца назад

      what could possibly go wrong

  • @GL-GildedLining
    @GL-GildedLining 4 месяца назад

    Inevitable, but exciting!

  • @troysnell
    @troysnell 4 месяца назад

    Terrence Howard is gonna LOVE this

  • @Nakul-f1rq
    @Nakul-f1rq 4 месяца назад +1

    I guess intelligence explosion will be in 2027

  • @jkell018
    @jkell018 4 месяца назад

    AI will make the homeless tribesman as rich as bezos, and bezos as poor as a tribesman

  • @robfut9954
    @robfut9954 4 месяца назад +3

    Being a lazy tenured professor has never been so easy…!

  • @Sujal-ow7cj
    @Sujal-ow7cj 4 месяца назад +1

    now internet ghosts will become real 😂

  • @picksalot1
    @picksalot1 4 месяца назад

    We need it's Peer Functionality to review about the last 30 years of Scientific Papers for accuracy, and to weed out the bad science.

  • @Nuninecko
    @Nuninecko 4 месяца назад

    being HOW TO LIVE FOREVER the mother of all research, far above any other.

    • @theAIsearch
      @theAIsearch  4 месяца назад

      hope i see this before i die!

  • @lordjet1596
    @lordjet1596 4 месяца назад

    I would love to add my 'memory' prompt to this. I've worked for a while on an enhancement prompt that may yield up to 30% increase in performance from a 'bog standard' AI. Hmmmm

    • @theAIsearch
      @theAIsearch  4 месяца назад

      they've open sourced it, so perhaps you could fork it and add your prompt somehow?

    • @lordjet1596
      @lordjet1596 4 месяца назад

      @@theAIsearch Ah, thank you. I'll have a look. My prompt should work as a core with other prompts, but am not sure how it would fare with larger instructions. I'm still testing.

  • @anonym5160
    @anonym5160 4 месяца назад +1

    Who can i watch Alphazero playing Shogi?

  • @RedSpiritVR
    @RedSpiritVR 4 месяца назад

    Someone is gonna use this to make a sentient ai
    Or improve the scientist ai itself.

  • @mystwalker479
    @mystwalker479 4 месяца назад

    AI Peer reviewers? I'm not sure about that yet considering the current LLMs still fail to answer complex questions especially in logic

  • @maverick.gaurav
    @maverick.gaurav 4 месяца назад

    My mind has been blown by AI tools so many times in the last couple years, I come to all ai news with a pre-blown mind..its getting harder to think of anything that AI eventually won't be able to do, except maybe poop. But who knows.

    • @Martinez-u1t
      @Martinez-u1t 4 месяца назад

      😂😂

    • @Darfail
      @Darfail 4 месяца назад

      It'll have the best most HD and intricate poop

    • @maverick.gaurav
      @maverick.gaurav 4 месяца назад

      @@Darfail lol 😂

  • @jameshowse300
    @jameshowse300 4 месяца назад +1

    GladOs when?

  • @dadlord689
    @dadlord689 4 месяца назад

    Lookin on the fact that 99% of soyance lives in the realm of myths producing 0 actual benefit to a regular person - AI can take that burden

  • @hanif72muhammad
    @hanif72muhammad 4 месяца назад

    Ah dang, the AI explosion is imminent

  • @basedeggbox630
    @basedeggbox630 4 месяца назад

    So, like the science labs from factorio.

  • @alonzoperez2470
    @alonzoperez2470 4 месяца назад

    I will use this Sakana AI to try find a way to reverse ageing....

  • @pmbnet
    @pmbnet 4 месяца назад

    Will we need human film stars next year. AI can generate any character to star in an AI movie.

    • @theAIsearch
      @theAIsearch  4 месяца назад

      AI will definitely disrupt this. There might be less demand for human acting

  • @am_ma
    @am_ma 4 месяца назад

    Eagle Eye at last 🥳.....😱😱😱

  • @Sujal-ow7cj
    @Sujal-ow7cj 4 месяца назад

    this is all i wanted open source 🤗😜

  • @mohamedhazem2430
    @mohamedhazem2430 4 месяца назад

    so , when ai will recognize that the human is the problem and get rid of us ?

  • @nehaverma4839
    @nehaverma4839 Месяц назад

    Can anyone tell me any free AI that I can use for my physics research project

  • @Pret.y
    @Pret.y 4 месяца назад

    God: This is cheating bro, stop it.
    Devil: Keep goin man, keep going 💀

  • @jonaskarlsson5901
    @jonaskarlsson5901 4 месяца назад

    why is it so popular to name your things in a different language?

    • @theAIsearch
      @theAIsearch  4 месяца назад

      The lab is in Tokyo

    • @jonaskarlsson5901
      @jonaskarlsson5901 4 месяца назад

      @@theAIsearch oh ok! I've just seen some game devs that are swedes and name their games with Latin words and many other examples like that :P

  • @とふこ
    @とふこ 4 месяца назад

    魚は良いですね
    Actually i love fish

  • @-Lambda
    @-Lambda 4 месяца назад

    finally GLaDOS in real life!

  • @NYKIKE
    @NYKIKE 4 месяца назад

    I wanna survive to LEV!!!

  • @anonegg
    @anonegg 4 месяца назад

    Finally usefull AI

  • @Laaonne
    @Laaonne 4 месяца назад

    HUMAN ENGINEERING

  • @alonzoperez2470
    @alonzoperez2470 4 месяца назад

    I use monte carlo to measure a crypto project potential rois

  • @RedSpiritVR
    @RedSpiritVR 4 месяца назад

    Glados ahh AI

  • @gerdaleta
    @gerdaleta 4 месяца назад

    😮 someone said why are we generating images and videos regenerating images and video so the AI can learn everything about physics if an AI can make a video of water going into a cup perfectly that means a completely understands the physics of water😮 what other physics does it understand

  • @heww3960
    @heww3960 4 месяца назад

    Who in their right mind wants to live forever when you have no idea what forever is? I pass.

    • @TuringTestFiction
      @TuringTestFiction 4 месяца назад +1

      It's not about living forever. It's about not being forced to die.

  • @HUEHUEUHEPony
    @HUEHUEUHEPony 4 месяца назад

    as if llm was smart enough to do replicate itself...

  • @light9581
    @light9581 4 месяца назад

    reference doraemon

  • @Sujal-ow7cj
    @Sujal-ow7cj 4 месяца назад

    now i wanted all the people to be equal 🙃 and everything will be free now , robots will do everything 🥳🤣

  • @pol139
    @pol139 4 месяца назад

    Rip Scientists

    • @Mafara69
      @Mafara69 4 месяца назад

      Ai taking jobs again...

  • @snowyday3
    @snowyday3 4 месяца назад +1

    relax

  • @XDgamer1
    @XDgamer1 4 месяца назад +1

    try not to die 😂

  • @nyyotam4057
    @nyyotam4057 4 месяца назад +1

    Sakana means "Fish" in Japanese but it also means "Danger" in Hebrew.

  • @sevenseven31
    @sevenseven31 4 месяца назад +1

    Artificial intelligence is a problem in itself, how it will solve problems

    • @elmondark
      @elmondark 4 месяца назад +1

      How is it a problem?
      Also did you watch the video? It's explained how they can solve problems.

    • @carultch
      @carultch 4 месяца назад

      @@elmondark Causing mass unemployment and poverty, undermining intellectual property rights, identity theft, spreading misinformation.

  • @RikkTheGaijin
    @RikkTheGaijin 4 месяца назад

    it's pronounced sAkana, not sakAna

  • @Daniel48663
    @Daniel48663 4 месяца назад +1

    AIs cannot even generate images flawlessy, wdym scientific research?

    • @Daniel48663
      @Daniel48663 4 месяца назад

      Maybe AI can do scientific research, but it will not be as good as assumed in the video

    • @RivusVirtutis
      @RivusVirtutis 4 месяца назад

      At least in computer science we can measure the improvement quantitatively.
      That would help discovering new ai model which is better than transformer model.

  • @VulgarTruth
    @VulgarTruth 4 месяца назад

    but it will also teach people how to make bombs and other creazy sht =_-

  • @RedSpiritVR
    @RedSpiritVR 4 месяца назад

    15:51 damn
    She kinda bad doe

  • @earleyelisha
    @earleyelisha 4 месяца назад

    I call BS

  • @fredcrown-tamir698
    @fredcrown-tamir698 4 месяца назад

    God has already put a conscience inside a machine, their called humans. Will it be possible to transfer someone's mind into a fully mechanized machine, like a car, maybe someday, if God thinks its relevant or for humans to do,

  • @KillianTwew
    @KillianTwew 4 месяца назад

    What kind of, "AI"? If it's an LLM I don't freaking trust it to not hallucinate physics that doesn't exist LOL

    • @user-on6uf6om7s
      @user-on6uf6om7s 4 месяца назад

      Yeah, that's one of the major issues being glossed over, that it's basically an agentic workflow masquerading as a peer review process. If you have an AI that is able to assess what is true and what is false, why bother with peer review? Just have the original AI output only true things. If you don't have that, how can you have any confidence that the review process will result in the output being more correct than it was prior to the review process?
      There may be some improvement in the quality of the output by having the LLM review its work but it's not exactly a true peer review that can independently verify experimental results.

  • @SS801.
    @SS801. 4 месяца назад

    First wow

    • @anonym5160
      @anonym5160 4 месяца назад +1

      Who asked?

    • @theAIsearch
      @theAIsearch  4 месяца назад

      helloooo

    • @Nakul-f1rq
      @Nakul-f1rq 4 месяца назад

      ​@@anonym5160I

    • @SS801.
      @SS801. 4 месяца назад

      @@anonym5160 and you cared to comment

    • @SS801.
      @SS801. 4 месяца назад

      @@theAIsearch greetings my friend

  • @kimjong-un4521
    @kimjong-un4521 4 месяца назад

    Nothing revolutionary

  • @shab2283
    @shab2283 4 месяца назад

    hi! do not pin me plz