CNBC: these tech industry deep dives are GOLD. Addictive and we're all rapt with curiosity. Keep it up, expand, and promote the writers and design/production staff.
It's fascinating to see Amazon making its move into the generative AI space with custom chips. AWS' dominant position in the cloud market could indeed give them an edge, enabling existing customers to explore generative AI easily. The competition is heating up, and it's exciting to see how this will shape the future of AI development.
I agree Amazon definitely wants to continue to raise the bar from a competitive and also from innovation standpoint. Which is expected to be a branch of an already established billion dollar industry. chip manufacturing.
Microsoft Azure is catching up and fast. It's likely to overtake AWS in the near future. In addition, Microsoft has plans to incorporate ChatGPT into most of its PaaS, IsaaS, and SaaS offerings.
0:10: 💻 Amazon is designing its own custom microchips to power data centers and AI models, positioning itself to supply the growing demand for chips in the industry. 3:17: 💻 AWS is making strides in generative AI and custom chips to catch up with Google and Microsoft. 5:36: ⚡ Trainium provides a significant improvement in price-performance for training machine learning models on AWS, but Nvidia's GPUs are still the preferred choice. 8:38: ⏰ Amazon is expanding its AI offerings, including large language models, and focusing on tools rather than building a ChatGPT competitor. 11:15: 🚀 Amazon is leveraging its existing customer base to drive adoption of generative AI solutions. 13:55: 🏁 The development of generative AI applications and the race to advance AI technology are ongoing, with concerns around security, regulation, and national security. Recap by Tammy AI
Using his analogy, since the racers get exponentially faster with every step they take, it is reasonable to consider who is winning within the first few steps.
This is an absolute delight. I had the privilege of reading something similar, and it was an absolute delight. "Mastering AWS: A Software Engineers Guide" by Nathan Vale
With this next gen' AI, everyone will have to become an entrepreneur. Work is over. They will need to own or access all the factors of production: -Land -labour - capital machinery - entrepreneurship Given that machines can also be labour being a worker in the strictest sense will become more and more unlikely because more and more of the workforce will be machines (of software agents). As long as society enables workers to become owners and or managers of the factors of production then I see all being well . If not, then society will become even more polarised between the haves and the have nots. In this case I see workers not voting for Christmas and Governments having to ban these next gen AI technologies all over the world.
I am a bit surprised that Amazon is not using the hardware from start-up Cerebras (CS-2), Wafer Scale Engine which seems to have some impressive proporties. Also, I have a hunch that some yet missing (not mature) building blocks like Non-Volatile Memory (NVM) like MRAM (VG-SOT-MRAM) would dramatically increase AI efficiency : therefore, big tech companies should actually collectively invest in scaling up manufacturing of this technology to remove some bottlenecks to enable new opportunities !!!
Its crazy that amazon is trying to bust into the extremely hard area of computer chips. But with 5.2B in revenue, they sure have a huge head start but it's going to be impossible for them to compete with intel any time soon. Since intel sports a nice 80B in revenue and already has the state of the art manufacturing facilities and decades of experience. Im sure if amazon wanted they could do it. But dont expect Intel to give up any market share anytime soon.
At the end of the day these guys want to sell products (whether most need them or not) and shove it down most people’s throats for good old-fashioned profits!All the wide-eyed jargon-dropping,awe-inducing interviewees are doing that,their bonuses depend on it.
But nobody able to make general-purpose microprocessors like Intel and AMD does. Even Apple's M1 and M2 chips also do not support many AI libraries and ecosystems. Different AI chips or just GPU would go haywire and divide the market and eventually the whole development system collapse.
Right now Nvidia is not meeting capacity making something like 400k H100's a year. There is quite a bit of room for other players but where they will really need to catch up is with software, AMD finally figured this out and at this point I'd guess everyone else is working on it as well. So long story short assuming there is no 'AI crash" it will probably be at the very least a few years before the market evens out.
China invented gunpowder, but fell victim to efficient European cannons. Americans are racing to develop and profit from AI, but the US could end up being sunk by this very technology. More AI will not mean increased happiness, but the deepening of political instability caused by unemployment. With the rise, success and hegemony of AI there will be a disincentive to laborious university studies and a consequent sustainable cognitive decline of the majority of the population. In the long term, the decrease in corporate profits caused by the reduction in the consumption of goods and services caused by the impoverishment of ever-increasing sections of the American population is predictable.
The only thing surprising about this is how late AWS is moving into custom chip design, if that is in fact the case. Perhaps the only thing new here is concerning AI custom chips, which every big player is surely doing by now. Tesla has been designing their own AI-focused processors for years already. It's just the obvious next step. The advantage of GPU's is their speed and throughput, they aren't specialized for AI. AI has unique needs which aren't served well by standard hardware.
We have a housing crisis and we are about to have food shortages, but at least Alexa will talk to us while we starve. I am amazed such a massive capital misallocation can occur after the Fed has hiked so many times. Rates are going to have to go to 20% to fix this nonsense.
Electroplating is easiest and cheapest to custom chips by drawing circuit on the plastic of other pcb then soak in electroplating aqueous layer by layer then any micro controller for sensor or anything control customize chip need but crystal processing light is much better using progressive lense and process light interference then very good if include transient lense on top then CPU easier
Amazon is incredibly ambitious when it comes to AI. When the fuss has died down with ChatGPT, they will be riding bigger wave. Their AWS employees must be being incentivised to a very high level. Whether all this will pan out to make them the AI leader, who knows.
What chips? Potato chips? People are using very normal clothing, no dust control environment? No wonder Amazon A.I. stuff fails badly, Nvidia built full dust controlled environment before starting chip R&D.
I think we all need to go back to business 101. What is the business case for my company using AI or in this case generative AI and LLM's. We are currently stuck in the tech and not the business. We always need to start from the business need since LLMs must have been used to try to solve some initial business need. If not then we are all trying to give a ROI to lab rat's AI discovery and make his pet project viable because they sunk in $1bln. This is not my problem or risk. This could be another Blockchain /Smart Contracts debacle. I hope you realise NO ONE is talking about Blockchain 😕😒😂😂😂😂😂
Rooting for AWS and anyone but NVidia - Nvidia can only charge monopoly prices for so long until custom chips can do the same work load for half the price
😂😂 all these database and Digitized leading more and more power need which leading huge requirements for Coal fired power plants leading to catastrophic leavel of global warming and US is facing such in large scale Devastatingly Wild Forest Fire and drought
CNBC: these tech industry deep dives are GOLD. Addictive and we're all rapt with curiosity. Keep it up, expand, and promote the writers and design/production staff.
this isn't a deep dive. they are telling you this because AI is going to put every company out of business in like 5 years.
It's fascinating to see Amazon making its move into the generative AI space with custom chips.
AWS' dominant position in the cloud market could indeed give them an edge, enabling existing customers to explore generative AI easily.
The competition is heating up, and it's exciting to see how this will shape the future of AI development.
I agree Amazon definitely wants to continue to raise the bar from a competitive and also from innovation standpoint. Which is expected to be a branch of an already established billion dollar industry. chip manufacturing.
😂😂😂amazon corpo propo
🎉
Microsoft Azure is catching up and fast. It's likely to overtake AWS in the near future. In addition, Microsoft has plans to incorporate ChatGPT into most of its PaaS, IsaaS, and SaaS offerings.
0:10: 💻 Amazon is designing its own custom microchips to power data centers and AI models, positioning itself to supply the growing demand for chips in the industry.
3:17: 💻 AWS is making strides in generative AI and custom chips to catch up with Google and Microsoft.
5:36: ⚡ Trainium provides a significant improvement in price-performance for training machine learning models on AWS, but Nvidia's GPUs are still the preferred choice.
8:38: ⏰ Amazon is expanding its AI offerings, including large language models, and focusing on tools rather than building a ChatGPT competitor.
11:15: 🚀 Amazon is leveraging its existing customer base to drive adoption of generative AI solutions.
13:55: 🏁 The development of generative AI applications and the race to advance AI technology are ongoing, with concerns around security, regulation, and national security.
Recap by Tammy AI
Thank you
The hostess is lovely. She also covered GPU market, Nvidia, etc.
Using his analogy, since the racers get exponentially faster with every step they take, it is reasonable to consider who is winning within the first few steps.
But different players are going to get bigger and faster at different speeds and some who are moving quickly now may slow down.
This is an absolute delight. I had the privilege of reading something similar, and it was an absolute delight. "Mastering AWS: A Software Engineers Guide" by Nathan Vale
Micro chip é a evolução. Futuro!
wow what a report. That was well done.
Better get SoCs included all with Generative AI and NPUs.
With this next gen' AI, everyone will have to become an entrepreneur. Work is over.
They will need to own or access all the factors of production:
-Land
-labour
- capital machinery
- entrepreneurship
Given that machines can also be labour being a worker in the strictest sense will become more and more unlikely because more and more of the workforce will be machines (of software agents).
As long as society enables workers to become owners and or managers of the factors of production then I see all being well .
If not, then society will become even more polarised between the haves and the have nots. In this case I see workers not voting for Christmas and Governments having to ban these next gen AI technologies all over the world.
Nope, Generative AI fails to do way too many physical tasks or write truly original papers itself
Should look at the nvidia clips of the chip designer tools. Gonna be a revolutionary future. Lets make some T100s baby!
Thanks, I am learning slowly I hope I will reach there this is something I have been looking for I just need more about this from you
Amazing! 👏
I am a bit surprised that Amazon is not using the hardware from start-up Cerebras (CS-2), Wafer Scale Engine which seems to have some impressive proporties.
Also, I have a hunch that some yet missing (not mature) building blocks like Non-Volatile Memory (NVM) like MRAM (VG-SOT-MRAM) would dramatically increase AI efficiency : therefore, big tech companies should actually collectively invest in scaling up manufacturing of this technology to remove some bottlenecks to enable new opportunities !!!
Amazon bASICs?
1000 IQ comment
😂😂😂
Just what we need; tech giants taking over even more of our lives...
Its crazy that amazon is trying to bust into the extremely hard area of computer chips. But with 5.2B in revenue, they sure have a huge head start but it's going to be impossible for them to compete with intel any time soon. Since intel sports a nice 80B in revenue and already has the state of the art manufacturing facilities and decades of experience. Im sure if amazon wanted they could do it. But dont expect Intel to give up any market share anytime soon.
How are their 8-bit Float (FP8) tensor cores? I have a lot of matmul to do.
2:05 thats aging like fine cheese although they have sniped the CEO.
At the end of the day these guys want to sell products (whether most need them or not) and shove it down most people’s throats for good old-fashioned profits!All the wide-eyed
jargon-dropping,awe-inducing interviewees are doing that,their bonuses depend on it.
it’s fascinating but at same time it feels its bubble which ought to burst
'How Amazon fails also on second attempt to deliver my new freezer, even after talking to customer service after they messed up the first time.'
But nobody able to make general-purpose microprocessors like Intel and AMD does. Even Apple's M1 and M2 chips also do not support many AI libraries and ecosystems. Different AI chips or just GPU would go haywire and divide the market and eventually the whole development system collapse.
It’s like we’re really watching the future unfold.
Here comes skynet !
FIRED.
Katie is beyond beautiful!!!
Rooting for Amazon
Hold up we about to have a crazy chip glut coming???
Right now Nvidia is not meeting capacity making something like 400k H100's a year. There is quite a bit of room for other players but where they will really need to catch up is with software, AMD finally figured this out and at this point I'd guess everyone else is working on it as well. So long story short assuming there is no 'AI crash" it will probably be at the very least a few years before the market evens out.
Thank u.
Pandoras Box.
Best wishes...
Thanks info
tweezers, soldering irons, and a freaking pencil!
China invented gunpowder, but fell victim to efficient European cannons. Americans are racing to develop and profit from AI, but the US could end up being sunk by this very technology.
More AI will not mean increased happiness, but the deepening of political instability caused by unemployment. With the rise, success and hegemony of AI there will be a disincentive to laborious university studies and a consequent sustainable cognitive decline of the majority of the population. In the long term, the decrease in corporate profits caused by the reduction in the consumption of goods and services caused by the impoverishment of ever-increasing sections of the American population is predictable.
Amazon could use IBM optical circuit technology
This feels like an ad ngl
There is no way they manufacture silicone chips, I bet they are using intel or tsmc, and they won't be using the latest 3nm process.
Agree, since silicone is for breast implants.
LLMs in the current form won’t be the state of art an year from now.
Uma tecnologia fascinante
Cool
what's the music in the end of the video ?
Red Forman was right Sodering is the future
The only thing surprising about this is how late AWS is moving into custom chip design, if that is in fact the case. Perhaps the only thing new here is concerning AI custom chips, which every big player is surely doing by now. Tesla has been designing their own AI-focused processors for years already. It's just the obvious next step. The advantage of GPU's is their speed and throughput, they aren't specialized for AI. AI has unique needs which aren't served well by standard hardware.
We have a housing crisis and we are about to have food shortages, but at least Alexa will talk to us while we starve. I am amazed such a massive capital misallocation can occur after the Fed has hiked so many times. Rates are going to have to go to 20% to fix this nonsense.
Electroplating is easiest and cheapest to custom chips by drawing circuit on the plastic of other pcb then soak in electroplating aqueous layer by layer then any micro controller for sensor or anything control customize chip need but crystal processing light is much better using progressive lense and process light interference then very good if include transient lense on top then CPU easier
Annapurna from Isreal
WOW>>>>
Imagine making a 3nm chip by hand.
yea this sounds so ridiculous. comparing TSMC multi billion dollar machine parks with an amazon office.
Nice Amazon has the Stability Ai its game over...
"Hey, we are building next generation AI chips"
"Oh right on, where can I drop an application?"
"In Texas..."
"Never-mind"
🎉
Hard to compete with AMD. Even Nvidia seem to be lagging behind AMD which is having advantage with Chiplets architecture.
Yeah so in 2 small rooms,a handfull of employees design chips. Fried chips
BUY #KUIPER 🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥
More like my garage "clean room" than tsmc's fabulous fabs of $20 billion
Amazon needs to be broken up.
Nice but Nvidia is too big for Amazon to catch up but good luck to how far they go.
Chip designing is one part and manufacturing is another .
@@superfluous5162That’s what TSMC is for. Amazon just has to design a good product and contract TSMC to manufacture.
OOOUUUUUU AMAZON SCARY
AMAZING 🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏
So amazon is making chips in the garage, by hand? What?
Even potato chip making is more complicated than that!
Design, not make. Watch the video again.
Amazon hq at 1999
Tony Stark made one in a cave. From a box of scraps.
@@youtubesucks8024 lol damn I forgot about that
Annapurna ❤
This makes me want to work for Amazon even more now ..
And get piped for no reason😢
I for one welcome our new AI overlords.
COOL 😎😎😎😎😎😎😎😎😎😎😎
Amazon ready to conquer chip market
Just call them what it is SKYNET..
When someone claims they are the best be suspicious.
How much did Amazon pay you for this video?
Annapurna - is it indian😮
Microsoft OpenAI ChatGPT is way far ahead against their competitors in the AI world. No doubt about that.
AI Software, not AI Hardware.
In which ways
@@mimi21746Agreed. Dependency on Nvidia is a weakness. Nvidia announced GH200 3 months after H100 bc of H100’s memory limitation!
Not only that but Microsoft Azure has doubled its market share of cloud computing in the last 4 years.
Amazon is incredibly ambitious when it comes to AI. When the fuss has died down with ChatGPT, they will be riding bigger wave. Their AWS employees must be being incentivised to a very high level. Whether all this will pan out to make them the AI leader, who knows.
Why does this feel like a commercial?
Evolucao
Ah yes Skynet is coming to help us with our human problem.
Yeah nice Amazon AD, nothing else. In reality MS is way ahead with Azure
Um
Databricks and Snowflakes eating big part of the cake
It's cool technology but there is no way anything with advanced AI will end up being good
People like Swami Sivasubramanian need to stop vocal frying when giving interviews because it is very hard to make out what they are saying 😠
Tesz robię czipy jak oni ale z gliny nie z papieru....
Everyone is Indian here
Annapurna is a Hindu name. Pronunciation is wrong. I think that Israeli startup has Hindu connection.
LOOK #ASEAN MARKET NOW !!!!
Those Amazon chips look like AMD chips!!! I smell a lawsuit brewing... ROFL
If you’ve been watching the Terminator movies over the last several decades, then you know where all this is really headed.
What chips? Potato chips? People are using very normal clothing, no dust control environment? No wonder Amazon A.I. stuff fails badly, Nvidia built full dust controlled environment before starting chip R&D.
Same observation here...Not a clean room environment whether for chip making or design?
I think we all need to go back to business 101. What is the business case for my company using AI or in this case generative AI and LLM's.
We are currently stuck in the tech and not the business.
We always need to start from the business need since LLMs must have been used to try to solve some initial business need.
If not then we are all trying to give a ROI to lab rat's AI discovery and make his pet project viable because they sunk in $1bln.
This is not my problem or risk.
This could be another Blockchain /Smart Contracts debacle.
I hope you realise NO ONE is talking about Blockchain 😕😒😂😂😂😂😂
So.. when are they going to start talking about the big-big savings the average consumer will get
The market is geared toward the upper middle class; go figure.
😂
@@sr4087 It's one of those things we have to deal with it
What does this have to do with the “average consumer”? This is for AWS which isn’t used by any consumers. It’s business to business sales.
Co to kvrwa za karp...
Amazon trying hard to stay relevant. Cute ad.
Rooting for AWS and anyone but NVidia - Nvidia can only charge monopoly prices for so long until custom chips can do the same work load for half the price
I am rooting for Azure! Their cloud offerings are amazing!
Why does the narration seem AI generated???
Jak kvrwa make... Jak oni je robią lurownica kvrwa w 5 nm... Ja pierdole... I to ma być legitne...
😂
Using cloud services as a weapon hurts trust and self-destructs. Your cutting edge tech won't help on long run. NOT TOO BRIGHT, t=886
Why are you selling alpha. Where is my money.
Why do half the comments in here seem like ai generated comments? 😂
What the guy said at 6:07 is true. He knows. We don’t need more custom hardware. All these people at aws look dead 💀
😂😂 all these database and Digitized leading more and more power need which leading huge requirements for Coal fired power plants leading to catastrophic leavel of global warming and US is facing such in large scale Devastatingly Wild Forest Fire and drought
No mention of Tesla's custom chip AI training chip or inference chip?