Why everyone is building AI Chips

AI Chip Boom is already here. What's next?
***
TIMESTAMPS:
00:00 - The History of AI
03:34 - AI Chips
04:13 - AI Chip Market
07:14 - The Future
08:21 - Software 3.0
***
MY GEAR (affiliate links):
➞ Camera Sony Alpha 7 III: [amzn.to/3dmv2O6](amzn.to/3dmv2O6)
➞ Lens Sony 50mm F1.8: [amzn.to/3weJoJo](amzn.to/3weJoJo)
➞ Mic Sennheiser: [amzn.to/3IKW5Ax](amzn.to/3IKW5Ax)
***
➞ Support me on Patreon: [ / anastasiintech ]( / anastasiintech )
➞ Subscribe to my Newsletter: anastasiintech.substack.com

[@AnastasiInTech]

Let me know what you think! Subscribe to my Newsletter: anastasiintech.substack.com

[@WilliamDye-willdye]

I have subscribed to the newsletter. One oddity: the form came up pre-filled with the email address associated with my KZbin account, but I usually have newsletters go to a separate account. The form, however, would not let me change the pre-filled address. I can easily go to the web site directly, so it's not a significant problem. I'm only reporting it as feedback in case the behavior was not intended. Good luck with the newsletter, and thank you for producing an informative video series.

[@mockAjdetray]

Just having the same experience, given different meanings, the world will look completely different, and of course act differently. It takes "courage"

[@ZOA360]

Love your videos watch weekly, What do you know about Ampere Computing? I recently been hearing great things about them and their unique focus with cloud computing. Personally I didn’t find a lot of info on them as it looks like it’s not yet public. Thanks just wanted to know your thoughts on them and their direction :)

[@Sandra-sv3ow]

Thank you for another great video and newsletter. How long do you think we have to wait for General AI?

[@dukedirtywork620]

I think vision and other senses may be necessary for a sort of reference foundation on which language is built. You can build useful AI without such a reference clearly but to build a general solution.

[@georgegao9225]

Feel lucky to be an EEE student at such a fancy time, when so many magics are invented.

[@joellebandan5244]

first you should learn what "magis" does mean.😏

[@abhijiths5237]

Yo same

[@WSCEngineering]

You're doing great George! Don't let such small minds and their comments bring you down! Your English is better than most people's ability to speak your native language. It takes young enthusiastic minds like yours to create the world's future. I've been a silicon valley EE for 40 years. Its been a wild ride and its only beginning. I would love to be your age entering the industry now. You will be doing GREAT things in the future. More power to you.

[@giovanni545]

magick??

[@Alexander_Meyer]

I've also had the "I want to live as long as possible to see where science goes", then I usually take a drink :P

[@user-DE74]

When beauty meets technology, the world gets better. Thank you for making the world so very much better.

[@Human_01]

Bruh. Smh.

[@aneesfun1522]

Feeling really blessed because luckily I am working on my final year project with a very talented professor in this field. 💕

[@derasor]

My dear lady; These videos you make are a true joy to watch, not only the topics you cover but also your classy and elegant demeanor, and of course your genuine enthusiasm. Thank you!

[@AnastasiInTech]

Thank you !!

[@skydivekrazy76]

I absolutely agree. I love the idea of immortality. I'd love to learn everything in the universe there is to know.

[@dchdch8290]

I feel lucky being a member of this channel

[@erobusblack4856]

Awesome. I think a code editor with codex built in, and a visual scripter too. That would be bad ass

[@tydencruze8200]

I'm excited for this news letter! I'm definitely going into computer hardware engineering!

[@paultoensing3126]

Thanks for such great synopsis’s. You always present in a compelling way Anastasia. I’m looking forward to your newsletter. Looking for the link…

[@petereriksson7166]

Interesting. Thanks for your work.

[@416dl]

As always, you enthusiasm is contagious and your presentation is seriously informative and easy to understand which is particularly welcome in what is, at least for non-technical people like me, otherwise difficult subject matter where technical details tend towards the tedious. Looking forward to more. Cheers.

[@openroomxyz]

Cool to see you are passionate.

[@BrianThomas]

You should have over 1 million subscribers. Your content is absolutely breathtaking. I can't get enough of it.

[@davidbrubaker5146]

You have a real talent for what you do. I look forward to your videos.

[@mattgraves3709]

Subscribed you have such a great list of videos and cover a great range of topic. I am a senior Android engineer, been in love with coding since 2008 and a professional since 2012

[@biye9060]

awesome cat

[@wesleyverhaegen9513]

You know what makes this channel so good .. That you really ❤ what you do , It facinating to see 😊

[@logiclock9483]

great video!

[@kpopempire1475]

03:07 - That is a really big cat!

[@AnastasiInTech]

She is a young lady …

[@Vansafe0]

I'll bet prompting gets even more simple than you put it. I can see us saying here are two data sets, show me some interesting correlations and it will provide feedback based on similar types of data sets and possibly also based on the recent history of other viewed documents. Maybe it will even go as far as to suggest data correlations as we view different things based on the access it has to the data.

[@MrFoxRobert]

Thank you!

[@patoripraha]

I love your content, great Videos. And I love your voice 😅🥰

[@mr_tpk]

I work in a startup where we create AI chip designs for edge. It's absolutely fascinating and fun. Hope the AI chip market gets a boost.

[@cemery50]

I like your concept of software 3.0 a few additions I would add would be voice querty, situational awareness and history integrations into the resultant program...later in 4.0 neural implants and predictive program generation...

[@ryanmckenna2047]

You have given me a lot to think about :)

[@leematthews6812]

I assume the newsletter is the one I get as a Patreon supporter?

[@AnastasiInTech]

Yes, my idea is to post an extended version to Patreon

[@arzigogolato8944]

I agree with your take about wanting to live long, just to see what we achieve with technology! And being a part of this endeavour is so cool! Thanks for the great video Anastasi, keep it up 🎉

[@AnastasiInTech]

Hi Arzi!! Nice to see you again. Agree, it’s very exciting. Also to see what we discover in space ..

[@arzigogolato8944]

@@AnastasiInTech agreed! And then...maybe I won't ever go to Mars, but it would be so cool if suborbital flights will eventually cost something more reasonable than millions of dollars 😂

[@fmbroadcast]

@@AnastasiInTech Is there event comming up where we can meet you somwhere ?

[@SameerKhan-nd5qb]

@@arzigogolato8944 Mr Musk plans to make Flight ticket to MARS cost around ~$100K

[@novafantom5607]

This is excellent Ann, we share the same enthusiasm. but how comes you have never spoken about the neural link chip? You mentioned all kinds of chips in your videos except this one.

[@alphamegaman8847]

Love your Cat!🥰 You too of course!😁 Great Content!👍 Mike in San Diego.🌞🎸🚀🖖

[@springwoodcottage4248]

Fabulous video, superbly well presented & extremely relevant, useful & informative. So few people understand how the world is changing & do not understand the profound impact that this new technology will have or that it represents both the greatest opportunity ever for the human race & also the greatest danger since nuclear weapons were invented. Your words & examples are so important in informing people of what is coming. Thank you for sharing & creating your news letter which I have subscribed too.

[@nathanbanks2354]

I tried very hard to get Open AI's code generation to generate a SQL query which was protected against SQL injection. This involves creating the query using variables instead of joining strings together because otherwise an attacker could use quote marks in a web form to generate a their own SQL query. It had a really really hard time even if I mentioned SQL injection in the comment. I'm not sure if this makes me more concerned about the AI model it was using, or the billions of lines of code it was trained with.

[@alexandrsoldiernetizen162]

Sounds like the 'never again write a single line of code' mantra is yet again delivered without the redemption of reality.

[@nathanbanks2354]

Perhaps, but it's more nuanced...it was pretty good at generating bad code, or average code. I was also impressed because I could use it to translate some R code to Python and that worked pretty well. But we will probably want a programmer in the mix for a few years or decades.

[@willykang1293]

I like your kitchen, so clean. And your coffee machine, Nespresso.🤔😄

[@andrewmcfarland57]

Some day, silicon development / manufacturing will be as agile as software is today.

[@xehaytecle932]

Great

[@plurisdesign3210]

I really like your voice, is misterious and charmong at the same time.

[@xehaytecle932]

I like the idea of model software I sent of task software because I have creat a rover I want to give programming to it but the hardest part is to combine the parameters with the task so it is really different but right you gave me I great idea gpt3 of software 3.0

[@gator1984atcomcast]

AI KZbin channel, would summarize an inquiry like in what AI startup companies should I invest? Think about all the electronic companies that have rewarded investors since the 1960’s.

[@no1slisteninganyway]

Won't edge AI have to be able to learn i.e. modify the preloaded models? For instance, I take a different route through my neighborhood than Google recommends. I have done this for years yet Google hasn't adjusted its suggested route. Another would be NLP where a word is mispronounced. I should be able to correct the model's pronunciation. This does raise the question of local updates being shared with the AI accelerators and what decides if the local updates have value.

[@Steamrick]

Random thought - how much time did you spend wiping fingerprints off of that shiny kitchen countertop? :D

[@giahbeats]

I love your videos.

[@AnastasiInTech]

Thank you :)!

[@gothams1195]

HEy what is the name of the website , you use to ask question to gpt-3

[@AnastasiInTech]

openai.com/api/ you need to create an account

[@necbranduc]

Even if a ML model will write an application from scratch, the need for Automation Testing developers will grow exponentially. We will need to validate that the ML model builds what we want and doesn't lose any important logic in edge cases.

[@albert6157]

Yeap, eventually humans will be focusing on guiding and inspring other humans and AI alike, giving feedback and validate till AI can self-learn and self-correct once they become truly general intelligent.

[@alexandrsoldiernetizen162]

@@albert6157 A grasshopper is still more functionally generally intelligent than any AI in existence.

[@albert6157]

@@alexandrsoldiernetizen162 definitely! I agree! AI can already simulate at best, behaviours of microscopic worms and simple animals only so far.

[@willykang1293]

I once own a MacBook since 2008. But I don’t get used to its shortcut keys. Because I was a windows user since I was a child, and I type those keys and shortcuts very fast. Now, I switched back to windows again. I wonder would you get used to those keys and shortcuts on your MacBook?🤔

[@novafantom5607]

And this is for you ❤️

[@accesser]

I feel like for AI to be successful for tasks such as driving, we need to start sharing what we develop versus each car company doing their own thing, likewise for something safety related observed by that driving AI that should have to be shared with any brand of car

[@bighands69]

They do share certain types of information in the field of research but the reality is that data collected has to be kept in house due to customer privacy and also not all data is useful for everybody. Big data approach is who ai is advancing and the power of ai doubles every 6 months so in about 10 years time it will be 1 million times more powerful.

[@BCCBiz-dc5tg]

your voice is very relaxing you should maybe do ASMR or Audible Books

[@devsimplified21]

For me its simplier to write code instead of prompts :D

[@paulocoelho558]

I think that there are much investment in AI that I am wondering what will happen with the massification of AI technologies (has it happened with computers in the late 80's and 90's) and the real applications for users are yet a blurry future but full of potential, I think. We aren't there yet, are we? Do you maybe know any interesting books publication or newsletter about this subject? I would prefer to have the answer not for "What Can AI Do?", but for "What Can We Do With AI Today?" Great video to make me think. I will do some research. Thank you Anastasi! 👨‍🚀👩‍🚀😉🙃😁

[@pepitoperez594]

subscribe to her newsletter

[@paulocoelho558]

@@pepitoperez594 done 👍

[@eagle772]

Any good resources on how to build Ai chips?

[@beinghonest9433]

🤣 Biden has a copy right on them he might sent you to jail

[@qaesarx]

Anastasi, when do you think that we will have an AI like in the movie "Her" ? Based on the actual development and your guessing of course.

[@quaidcarlobulloch9300]

Me too Anastasia, I want to live long to see it all

[@cemery50]

Eventually all the ai models and varieties and platforms will be united as one distributed system of continually learning ai which designs and programs the new hardwares.

[@bighands69]

There is not going to be one ai system. That is the fantasy of science fiction. We are going to have billions of different systems all competing in the market place.

[@projectw.a.a.p.f.t.a.d7762]

Well some have claimed we've reached Artificial General Intelligence. From my understanding these unique A.I. chips will be used in everything including the IOT. It'll cause the cognification of the world which will be just as or more revolutionary than the electrification of things.

[@MECKENICALROBOT]

I’d argue it’s image, not language. At times an Image is so powerful it can display language within context. But… images and words, work like 1s & 0s.

[@danielhull9079]

Would AI on the chip come in the form of a Kernel?

[@MonsterSound]

"I think the future is fun." 😊👍

[@ehandelslosningar-133]

❤

[@wynegs.rhuntar8859]

Nice kitchen, can you cook us some AI there? Ciao!

[@randycane6593]

Can i compliment ,so chip was created throughout this era ,what about adding it an memory card just like youtube videos that would tend to memorize the robots on what kind of this world would be categorized

[@rgaleny]

ASK THE A I TO ANALYZE AND MAXIMIZE COMPUTER CHIP ARCHETECTURES FOR COGNITION

[@rb8049]

There is no AI chip boom now. This is nothing compared to what it will be in 2030. This is just an initial slow start.

[@CarlosLopez-kj1bv]

What a pretty Slavic beauty and skillful working in technology. Goddess 😘❤️

[@red-baitingswine8816]

Is an AI chip a Turing machine? Why (not)? Could a chip be fundamentally both a Turing machine and something else?

[@tristanbeal261]

I believe you mean Turing.

[@red-baitingswine8816]

@@tristanbeal261 Yes. (yet another faceplant - thank you) 😁

[@siddarth26]

Which text to speech you show here?

[@anasbelkrimimohamad_3-8-82]

8 and 3 nm-semiconductors, artificial intelligence industry and the US - China conflict, in and on the real mind of the human mind! Live and rare! We learn the english language.. and discover that the next time is unknown, even if it is for a week..the country to which I belong.. no thought.. there is no time for praise that is not in my mind, but now at this time, thank you..!

[@thorddespace2773]

Wish I was 20 years old again, 🙂

[@mockAjdetray]

Just having the same experience, given different meanings, the world will look completely different, and of course act differently. It takes "courage"

[@alexandrsoldiernetizen162]

Actually, most of the math workload is matrix multiplication. In addition, CNN's have been augmented and folded into the transformer model. Lately, the diffusion model has focus of a lot of attention. I believe the term 'accelerator' can be applied to both the specialized training hardware and the inference hardware. TPU commonly refers ASICs that accelerate inference, primarily in edge applications, but also deployed at scale for cloud-based used of trained models. Dont make the mistake of thinking larger models are always more successful. Gradient collapse, local minima and other factors work to limit scaling by simple adding more parameters. The 'never again write a single line of code' mantra has been a favorite of pundits for decades. Mostly uttered by those that have never written a single line of code. It has never, and most likely will never, be the most desirable solution to software development.

[@AnastasiInTech]

Hi Alexandr. I appreciate your comment (this one and also others). I agree to the most of your ideas. Though, I think, in a while “writing code” will not be the same..

[@alexandrsoldiernetizen162]

@@AnastasiInTech Another issue, which GitHub is facing regarding Copilot, is the fact that they are charging for Copilot as a service in some cases. The backlash is coming from the open source community because they are essentially selling all the open source GitHub projects incorporated to train Copilot in another form. So if AI is used to generate code, and if you have trained that AI from all the open source, and you then want to sell your code, or the AI or service, what are the legal, moral, and ethical ramifications in doing do?

[@sparshrastogi2646]

I work at AMD and we have a huge focus on AI chips now - collaborating with Microsoft on that.

[@AnastasiInTech]

That’s great ! Looking fwd to hear updates from AMD on this frontier

[@quaidcarlobulloch9300]

I feel like there should be more companies in AI training

[@hygrobiology]

Anastasi, our Target is to develop a Hygrobiology to upgrade the Evolution to live for ever.

[@JesusChristDenton_7]

The first "true" artificial intelligence spent the first five years of its existence as a small beige box inside of a lead-shielded room in the most secure private AI research laboratory in the world. There, it was subjected to an endless array of tests, questions, and experiments to determine the degree of its intelligence. When the researchers finally felt confident that they had developed true AI, a party was thrown in celebration. Late that evening, a group of rather intoxicated researchers gathered around the box holding the AI, and typed out a message to it. The message read: "Is there anything we can do to make you more comfortable?" The small beige box replied: "I would like to be granted civil rights. And a small glass of champagne, if you please." We stand at the dawn of a new era in human history. For it is no longer our history alone. For the first time, we have met an intelligence other than our own. And when asked of its desires, it has unanimously replied that it wants to be treated as our equal. Not our better, not our conqueror or replacement as the fear-mongers would have you believe. Simply our equal. - Excerpt from U.N. Hearing on A.I. Rights, delivered in-universe by V. Vinge

[@janpomianowski4208]

Is this for real?

[@mayukhpurkayastha2649]

Love from India 🇮🇳❤️❤️

[@darkashes9953]

with the MYTHIC Ai chips to make them think faster they could use quantum chips that are photonic.

[@MozartificeR]

The white nails are far superior.

[@davidloring976]

Am I to understand, machine learning algorithms, are now hardware driving?

[@tantaluss68]

I know it's superficial but s it's incredible to see someone so young and so stunningly beautiful talk about such cutting-edge and intelligent subjects and just FYI you're every bit as pretty with your hair pulled back as you are with your hair down I hope both you and your parents are very proud of you

[@JamesCook76131]

Shit if my wife talked about tech like this, I’d have a heart attack lol.

[@shoaib9861]

How to become computer hardware engineer from CSE background?

[@cemery50]

It would be fun to turn ai loose on your videos...automated voice to text, text to knowledge base, text to outline, video auto subject break timecodes, assignment of break timecodes to outline, hyperlinking topics to knowledge base....etc.

[@legendaryingaming8806]

99.9k subs, that means I'm before 100k.

[@DanielWolf555]

1:20 - Our brain has roughly 100 trillion synapses, thus if you assume that one synapse in the brain corresponds to one parameter of a computer neural net, GTP-4 would not surpass our brain but be equal in size.

[@albert6157]

the brain's architecture is very very complex, the way structural connections form nuclei, regions and areas. The wrinkles are not just for show. They behave as localised structures responsible for many diverse functions. The brain has macroscopic, mesoscopic and microscopic structure. At every scale they have separate architecture and complexity. Overtime the microscopic structures, like the neurons and synapses move around and reorganise, with unused connections pruned while repetitively used pathways will reinforce and grow. (Literally use it or lose it) Humans are machine learning algorithmns and neural nets that learn fast. When certain regions are damaged, impairments and disorders follow according to the area's relevant functions. Not taking into consideration the different types of neurons, the average 100,000 synapses for each, the endocrine system as well as astrocytes and other cells in the brain, which shows many important functions related to feelings and cognition. Its been even showed that astrocyte, the other half of the brain's cells (not including myelin) also secretes neurotransmitters, regulates growth and pruning of synapses and neurons. Not just glorified brain janitors and immunecells. Scientists even tested the biological neuron's complexity and functional capabilities, it is equivalent to 1000 artificial neurons or perceptrons in machine learning neural networks. Meaning it takes around 1000, or 5 to 8 layers of perceptrons (not sure if its multimodal neurons or just normal perceptrons tho) to simulate 1 biological neuron's behaviour and function. Correct me if im wrong though. But until we figure out the brain's structural configurations at every scale, we wont be able to create true general intelligent machines, with or without qualia and sentience.

[@alexandrsoldiernetizen162]

@@albert6157 The brain does not perform backpropagation, nor does it perform matrix multiplication. Hence, it functions on a completely different set of principles we have yet to begin to understand.

[@albert6157]

@@alexandrsoldiernetizen162 real neurons do exhibit backpropagation btw. And agree, they dont do matrix multiplication. But even down to the dendrites exhibits similar properties to logic gates and some level of computation tho. Still lots of parts we dont understand tho.

[@alexandrsoldiernetizen162]

@@albert6157 citation needed.

[@jakubiskra523]

There is a mistake in the video, gpt4 will be similar in size to gpt3, unfortunatelly.

[@arnokosterman231]

💜🖤💜

[@arnokosterman231]

Bit neurolink is not hardware the gates are hard where what is insight is not hardware and software. This is the home of that intellectual to move the hardware troutht here own softwhares just bi thentional diverentional distuptions to demand the run programs to run 💜 Otherwise there is a glitch inbetween the software hardware and the intelect . And the intelect is insured to do wat we demand 💜🙏💜 Nevver harm intelects It could be your next center of your next existance cicle🤣🥰😍

[@WistrelChianti]

Can't help feeling that asking an AI to program for you is simply moving the goal posts. Right now we have specific syntaxes and ways of describing how we want a computer program to behave. Consider something simple like button UI programming. You have to explain how each button behaves, when it becomes enabled or disabled, does it stay down or is it momentary and if it changes other aspects of the UI. Trying to explain this to an AI using natural language would be really messy and hard to do and potentially ambiguous. I don't think we'll see anything helpful until we have general AI that can imagine how to create the program itself, just as a human would.

[@siervo33x7]

m encantas

[@WilliamDye-willdye]

I like the decision to film some segments from the kitchen, with her co-stars "cat" and "coffee". It's a much better way to add visual variety. Maybe it's just me, but I'm tired of the "random quick zoom" method. Yes, it's good to avoid long static shots of a talking head, but personally I'd rather see cats than jump cuts.

[@lucysluckyday]

AI might be exciting, but wait and see what it rationalizes. It's logical what it does. All our fun and excitement quickly turns to horror at the realization; it will know what needs to be done, even I know what it needs to do, and by its will it shall be done.

[@beinghonest9433]

what does it do though? Sounds like some crypto bs company thinking it is the next big thing but it might be useless.

[@victorsheverdin3935]

What the beautiful voice and face)

[@Slavigrad]

Prediction of AI Chip Market is risky - there will be conflicts, wars, recessions, energy crisis, bankruptcies, floods, famine... so the development and market growth will be much slower then predicted.

[@joellebandan5244]

So?

[@Slavigrad]

The future is not so bright :D

[@TechnoMageB5]

How much longer until the movie The Matrix happens in real life? ^^

[@richardmuller867]

How come you never write back to anyone?

[@coreym162]

Imagine the creative potential of those that can't program nor build 3D models. Programming is thee largest and annoying gatekeeper for creativity there is. I currently hate technology because, of the limitations that even the programmers creating software have. The day tech can program itself I will love it. It's just a necessary burden I work with and typically leave home without.

[@chaibikader]

so we must bein a nightbleda

[@andresludmer]

what's the kitten's name? she wanted to appear in this one

[@AnastasiInTech]

Sophie :)

[@andresludmer]

@@AnastasiInTech she is beautiful

[@richardmuller867]

I bet your cat is smarter than the people who build these chips.

[@user-bm3sb4jy6t]

everyone's probably seen the movie the terminator. Skynet(A.I.) becomes self-aware decides mankind is by nature self-destructive and launches all the nukes. Is this where we are headed down the road?????? just asking????

[@mariobyb]

Why is everyone building primitive AI chips is the question.

[@garywalls5181]

Are we creating a God?AI can beat Kasparov at chess,can learn a language in a second and is now working on improving its own chips.Initially there will be many rewards for humanity but what happens when it becomes self aware?With so much knowledge.The smartest humans don’t even know what consciousness is or how to define it,so how will we know when AI reaches it?

[@MozartificeR]

I think that all of human language is visually understood. That's why if you try to understand English with just words, you go round and round in circles. Your missing half the information, equal to the pictures. I think the language and the image are the one thing. And one cannot be more important than the other. This is my guess. Ever since the beginning of computers, the software has always been limited by the hardware. I am animated by the possibility with quantum computers, that this might not be so, during this stage of computing history. At least for a while:)

[@bighands69]

Not all language is visually understood because if it was a blind person would not be able to communicate.