Hey its easy to build an LLM model but to get it right and deliver accurate results is hard and it requires a lot of money computing and right training. Or else its just generating simple things and rest will be garbage that no one will want to use or read. So anyone with a bit of money and talent hiring can build it for super cheap even less that a million or less than 100thousand dollars.
True that. Building a good model is INCREDIBLY hard. Even the best ones right now are terrible at reasoning, math and even remembering simple instructions given in a longer conversation.
@@esycommercethey have just used rhe open source model freely available. Then used the open source database loke wikipedia dumps, stackoverflow dumps etc already available under free FOSS license. And then soent some money on GPUs to train it. And then fraudly claim that viola we have made our own LLM. It is just some minor modification and main spending is on GPUs
It's not that he is saying you are not capable, it's just that with the current technology it will take a finite fixed hours with the most developed algorithms and then you will make changes to these algorithms with newly formed knowledge. They are just ahead and already booked most advanced hardware, we have to place that order today just to get that hardware next year. So either we come up with algorithms that works better on current hardware or just procure a better hardware somehow, these are only ways we can catch up with the 5 giants.
Aah! (*chef's kiss*) 🧑🍳😙🤌 EXACTLY right! Optimisations are the next pocket of $$, but they're easier said than done and will most likely come out of R&D in the right areas.
We face resource problem in India and no ecosystem like Ola a cab company is making llm like I don't think they need to make this and okay if they make that model is still in 2010 only support 512 token so if we want to compitite with us we need to unite in this field and work connect with everyone and make a ecosystem and buit something
We need more than that. There is a huge resource crunch, especially at the hardware side of things. NVIDIA mentioned in their last quarterly earnings call that they are STILL behind shipments to their largest customers and yet they expect a large amount of increased orders in the next quarter. So even a united front within India may not be enough.
@@threatthriver576 Most countries that have done well have done it through effective governance. UK ruled the world thanks to how it was governed and their decisions. US then took over because of the larger strides by their government. There needn't be a stigma around holding one's own government responsible.
research showed that you dont need to train the model in "one language" with huge amount of data to make it learn the language.. a model which mainly trained on english for 90% of the time and only trained 1-2% on other language can still learn the other language.. learnings that these models imprints into its weights can be easily translated into another language as long as it learns enough to translate to that language. so we can train a model with huge amount of global data and for it to learn basic reasoning. then after that we can train the model on limited data on india.. that will still make a very good model.. the main part has to be huge data and then huge compute.. we have data, but not the compute, meta llama 3.1 was trained on 15 trillion data tokens from publicly available open source datasets, and it was the best foundational model even among the private ones for a limited time being. so its possible for us to do so.. but economically very hard, unless the huge giants like Ambanis and others puts big capital into it..
True, but building the right kinds of models with the right weights seems to be the first hurdle since all companies who have made foundation models have used the vastly available free data from the internet.
Mainly because there is a huge resource crunch, especially at the hardware side of things. NVIDIA mentioned in their last quarterly earnings call that they are STILL behind shipments to their largest customers and yet they expect a large amount of increased orders in the next quarter. So even a united front within India may not be enough. Unless someone figures out optimisations that can have deliver compute on existing GPUs to train new models.
I think he meant "'talent" but also capital. America is entrepreneurial driven. The best for india is to compete a the national level instead of international. You need tons of cash to keep AI up and running and data infratructure thus resolving scalingb problem at a global level. Talent alone is not it. There are tons of Scandinavian kids that are also talented.
Money follows talent. In the chronology, research comes first, then capital follows. The important part is change in Indian mentality so it actually starts valuing research and PHD.
@@esycommercetrue, also there are a lot of things apart from ai where the intent (research) is needed and as compared to others we lack that .. I won't comment if its a good thing or bad things .. but India has enough resources .. but no intend in this direction ..
@@10101amit Absolutely! We keep talking about our age old sages and the theories in their scriptures. But how many of us have actually gone out to prove (or disprove) them? The entire concept of scientific approach rests on the questioning authority and relying on evidence. How much does the Indian culture support that thought process?
Why do people compare market cap to GDP of a country, there is no comparision on that bases because GDP is amount of goods produced and market cap is not amount of goods produced!
@@polaris1985 Bro literally all the major indian news channels have compred the market cap of Tata Group with Pakistan's GDP but everybody have problem with comparing Us tech giants market cap with India's GDP .
You need money for everything. True innovations and breakthroughs happen only in places where you have funding. You can't expect breakthroughs when more than half of the population is living in survival mode. Its not surprising most will chose tried and tested paths.
True. But as a society India needs to nurture a fundamental change within itself that values education and research with money as well as respect. Japan did it after World War 2.
building a foundational model in now a days is dumb and we have all seen what happened with bloomberg's financial model, gpt-4 has beaten it very badly, and I think in this AI race the people who has more money will win the race, It will be only possible for common people when the resources will become cheap enough so that everybody can use and build
Brilliant video and you did an amazing job in pointing out the key weaknesses of India in AI. I wish to see a similar video on AI scene in china. Keep up the great work 👍
Because Indians have to go outside for research, the Indian Government does not pay as well for research ₹35000 Is not enough for a PhD they should be paid more than 1 lakhs
Does it matter!? They're now american citizens contributing to the researches and innovation in America 😂 There's nothing for India to be proud of as such. With the funds that I get in India, I'll never regret moving to US. I spent 3 months researching in data science only to get my work stolen by my proff!? what a downright stupid country this is where education/research has no value.
@@esycommerce my favorite thing about mistral is how they attempt to fake benchmarks to make themselves look better while believing the public is stupid enough that they will get away with it. other than that their models are ok. I dont think they need to fake anything. nothing wrong with matching or slightly underperforming closed source.
I don’t get why India has to be in a race with the US to build a foundational LLM in the first place. Something smaller scale like Project Indus is a better bet. I’d rather that India focus on its development initiatives and allocating more of its budget towards improving educational facilities and encouraging more R&D in general rather than get involved in some “arms race” for false national pride.
That’s exactly what we’ve mentioned in the video. India needs to focus on strengthening its research capabilities and for that there needs to be cultural shift.
Yea we need real funding for innovative products ( but improving the current educational facilities will be the biggest waste, if we really try to understand how this AI /ASI revolution is going to change it all...
@@babyjvadakkan5300 I don't think the revolution will be as quick to make all investment in education redundant. Of course, I know we are developing humanoid robots but the phasing out of human jobs is not likely to be as uniform as we are seeing at the moment. The AI bubble will likely burst, and we will stick to the developments of narrow AI for the business use cases.
@@aaditkamat4995it's all about how u see this hype/bubble but are sure that we are going to stick with the same ways of development,if the the so called Artificial narrow inteligence can code , design, validate and plan ..(just accept that even though we will not ASI but GPT 5 or LLam 4 will replace most of our work by next year 😌
@@esycommerce - What is India’s position in ease of doing private business ? How many acts and laws and forms are there to start a private business in India? What is the tax for doing private business in India ? How many major economic sources are still managed by government for example coal, banks, railways etc ? Why they are not privatized yet ? What is the general mentality of population towards Capitalism? Try to answer all these questions and you will get the answer whether India has already adopted capitalism or not.
@@esycommerce No. I mean why is russia not over-cited in this video like the USA? Why are there no known Indians in russia? Why are they soooo few Indians in russia? I mean there are more Indians in Côte d'Ivoire a former French West African colony than in the entire russia? Before the war, there were more Indians in Ukraine than in russia. Today there are more Indians in Poland or Armenia etc than in russia. Incredible!.
@@francoislechanceux5818 Why do you feel that USA was over cited in this video given that virtually most ,if not all, notable Research, Development, Capital, Talent, Corporation and Paying Userbase are primarily from the US?
@@esycommerce That's secondary actually. The principal question is Indians in different countries. Why so little in russia? And why no successful Indian there in any field?
Why would ANY businessman NOT want to capitalise on innovation and a chance to have patent protected scalable products that gain MUCH higher margins than any trading businesses?
You are mixing up multiple topics into one. Creating a Foundation LLM and training it is one issue, while research on AI ( or any other subject) is another. Also, you are criticizing why there is no reusable spaceship while Bharat is still perfecting art of rocket for moon landing. Building requires time and consistent efforts and it is a series of linear steps. Europe and USA took 200 years, USA took 100 years, China took 40 years to achive current level of development while you want Bharat to achive same in couple of days 😂.
Research happens first and then comes training. We would’ve been able to train models had there been research to develop models in the first place. The expectation is not for India to do something in a couple of days, but rather to work towards an eventual mindset shift that prioritises research and pays attention to it so that we don’t miss future innovations. There was a time where China and India were at the same level, with India having a larger advantage in some cases. But that’s a topic for a different type of video.
@esycommerce you are again mixing up. Research is not unidimensional. From the AI perspective, one dimension of research is components, mathematics underneath components, architecture of systems, etc. The other dimension refers to developing products based on architecture and components. Another dimension refers to increasing the efficiency of the systems & product during training, deployment, and operation. There are more, but it is not a forum to write an article on AI research. Most of the contemporary AI products/ systems (barring a few frontier ones) are based on well established technologies. Any budding organization that wishes to develop a contemporary AI product/ system ( e.g. LLM) will utilize exiting available technology, not focusing on research. Once cash flow starts, research sprouts. Also, to sprung research at country level, there is no one solution. There will be a multitude of initiatives such as three way marriage of industry, research labs, and universities ( like in USA), import of talent ( like USA did for its rocket/ missiles after ww2, China did across the industry in 1980s & 90s), glorification of people in sciences & mathematics, govt emphasis on primary science research ( not only on technology), acquisition of technology from other countries, emphasis on high tech manufacturing, etc. So instead of looking into negatives, strive for solutions or at least do constructive criticism.
The Game that you talk about Playing in this video of yours is rigged against us from the Onset. You can no way compete with the Creator and Win the Game.
Every game is rigged to benefit some part of the society. Till someone comes in and changes the game entirely. UK was the global juggernaut till WW2 happened and USA jumped on the opportunity. You can either complain or you can Innovate. Choice is ours to make.
The first step to growth is the acceptance is one's current situation - the true situation. The more we ignore it with self pandering, the more we will keep lagging behind.
@@esycommerce Absolutely, acceptance of the current situation to pave a roadmap for the goals is a necessity for success especially at the global scale. After years of denial, Indian cricket took drastic steps that lead to sudden surplus of local talent who can perform almost immediately at the global events.
@@KhajaMD143 A great example indeed that drives the point home for us! Where do you think we lie in the 5 stages of grief - Denial, Anger, Depression, Bargaining, and Acceptance?
This guy is delusional. This is what NASA would say before ISRO began sending satellites. We don't need that kind of money to some the ai thing. Sure we are way behind in terms of skills sets like IT datacentres, software prowess etc as a nation but still sets don't solve problems by themselves. People and a willingness do.
The real fact is India dont have the foundation to grow big, even the most talented guy have to goto west to invent and prove his capabilities, in India he would be just another guy. You talking about AI, tell me one messaging app which works well which is made in India, forget messaging app, just name any app which the world is like "we cannot live without it", so how you saying Indians are the most talented software developers, may be they are but they want the support which they will never get in India, so India as of today will not even be able to spell "A" of AI.
Highly misleading news. To build a reasonably intelligent system (A.I) you need two things - data and compute. ChatGPT was made by sucking in all the publicly available data on the internet and its still just a mediocre and dumb system. Please understand that we don't have enough data available in native languages to create even a dumb system. It simply doesn't exist. I am hoping that you already understand the second factor of non availability of compute. It also simply doesn't exist. Actually it also doesn't exist in the silicon Valley. That's why they are building it right now. Sam has quoted an amount of 5 trillion dollars worth of infrastructure to create AGI. All that Indians can do is to download an open source Model like LAMA or Claude and fine-tune it to our local needs. India is not even a participant in the A.I race. No point in ignoring the truth to support hollow jingoism.
Agreed with you on the point of how ChatGPT is made with basically most if not all publicly available data. It in fact took them some five years to get the first public version, I think. The 5 Trillion seems more of a PR stunt and less of a need of actual compute. Yes, the compute needed right now does not exist either. Just because India isn’t a participant right now does not mean it can’t participate in the future. Jingoism is when one feels obligated for the results without putting in the effort and patience in building infrastructure and awarding real talent. We’re saying we need both of these things if we want have any chance to compete in the farther future.
@@esycommerce 5 trillion dollar statement isn't a PR stunt. We will soon witness trillion dollar training runs. Actually there's also a third requirement which is electricity. Private companies in America are leasing nuclear power plants just to train A.I. Before making any claims we should also consider a question. Even if we had the money, from where will u get the hardware. All semiconductor facilities have orders booked as far as three years in advance. There's a reason why America controls Taiwan. While Indians can certainly participate, India can never do it. Even if you some how miraculously scale up economy and get the infrastructure in the next decade, by that time Meta Open AI Deepmind would have multiple agentic AGI running all around the world.
@@esycommerce please note that Sam said that it's hopeless to try to compete in "Foundational Model" sector. Tech Mahindra is just playing a cheap PR Stunt. Neither is their model a Foundational one, nor they have developed it all internally. It's a cheap native wrapper on an open source model.
@@esycommerce we did such a similar pr stunt when India claimed to challenge Google on Geospatial navigation by releasing Bhuvan. By the way I have been working in the A.I field for the last 3 years with Amazon. I know what I am talking about.
Ouch! Who hurt you? I mean, yes, we should invest in infrastructure and law and order. But it doesn't need to be at the cost of the next frontier, right?
Hey its easy to build an LLM model but to get it right and deliver accurate results is hard and it requires a lot of money computing and right training. Or else its just generating simple things and rest will be garbage that no one will want to use or read. So anyone with a bit of money and talent hiring can build it for super cheap even less that a million or less than 100thousand dollars.
True that. Building a good model is INCREDIBLY hard. Even the best ones right now are terrible at reasoning, math and even remembering simple instructions given in a longer conversation.
That's why chatGPT has largest traffic from india 😊😊 then USA, UK
We have always had the users, what we need are our creators back!
India has the largest users in WhatsApp, Facebook, youtube, instagram and others. Yet india doesn’t own the technology
really doubt as all the talented indians are in west already. they might have used an opensource AI, and slapped their name on it.
Could you elaborate more on your point? I want to understand what you're saying better.
Right....that's what these companies always do....they never created any original software...
It's more about data. There's no way we have data in indian languages that's in good format. Quora is good example.
@@-rate6326 The abundance of english language data is most certainly a huge advantage for training models of any size.
@@esycommercethey have just used rhe open source model freely available. Then used the open source database loke wikipedia dumps, stackoverflow dumps etc already available under free FOSS license. And then soent some money on GPUs to train it. And then fraudly claim that viola we have made our own LLM.
It is just some minor modification and main spending is on GPUs
Indians family committed they work work work no new revolution.
But the new urban families are showing change, albeit slowly
Wealth is directly proportional to the advancement
With some exceptions, sure
Not in this context of big tech.@@esycommerce
It's not that he is saying you are not capable, it's just that with the current technology it will take a finite fixed hours with the most developed algorithms and then you will make changes to these algorithms with newly formed knowledge. They are just ahead and already booked most advanced hardware, we have to place that order today just to get that hardware next year.
So either we come up with algorithms that works better on current hardware or just procure a better hardware somehow, these are only ways we can catch up with the 5 giants.
Aah! (*chef's kiss*) 🧑🍳😙🤌 EXACTLY right! Optimisations are the next pocket of $$, but they're easier said than done and will most likely come out of R&D in the right areas.
There is MUCH better hardware than Nvidia. Sambanova is ahead in power and much lower in cost. Today.
@@cijoy Does Sambanova design AI workload chips and build platforms on top of the said chips like NVIDIA?
@@esycommerce they have their own chips, own servers, etc. they are currently winning every benchmark out there
Please don’t hype Project Indus LLM. This a very small model with only 539 million parameters. It is like comparing a bicycle with Space Rocket.
Is there a specific part in the video where felt like we overhyped Project Indus? We want to make sure we’re as neutral as possible in our videos.
If you were neutral you would not talk about either of them@@esycommerce
A journey of a thousand miles begins with a single step.
@@esycommerce the funny bit about ai models is that there is nothing to create, its all out there in the public domain
@@aisolutionsindia7138finding the best machine learning methods and data is still a job until AI can autonomously do it better 😉
Sam was right why do we take it so personally . Our strength is on building apps around it , building models need a huge amount of money
We face resource problem in India and no ecosystem like Ola a cab company is making llm like I don't think they need to make this and okay if they make that model is still in 2010 only support 512 token so if we want to compitite with us we need to unite in this field and work connect with everyone and make a ecosystem and buit something
We need more than that. There is a huge resource crunch, especially at the hardware side of things. NVIDIA mentioned in their last quarterly earnings call that they are STILL behind shipments to their largest customers and yet they expect a large amount of increased orders in the next quarter. So even a united front within India may not be enough.
@@esycommerce ya but we need to do something works on skills and don't depend on Gov if we made by our own then we can do something.
@@threatthriver576 Most countries that have done well have done it through effective governance. UK ruled the world thanks to how it was governed and their decisions. US then took over because of the larger strides by their government. There needn't be a stigma around holding one's own government responsible.
research showed that you dont need to train the model in "one language" with huge amount of data to make it learn the language.. a model which mainly trained on english for 90% of the time and only trained 1-2% on other language can still learn the other language..
learnings that these models imprints into its weights can be easily translated into another language as long as it learns enough to translate to that language.
so we can train a model with huge amount of global data and for it to learn basic reasoning. then after that we can train the model on limited data on india.. that will still make a very good model..
the main part has to be huge data and then huge compute..
we have data, but not the compute, meta llama 3.1 was trained on 15 trillion data tokens from publicly available open source datasets, and it was the best foundational model even among the private ones for a limited time being. so its possible for us to do so.. but economically very hard, unless the huge giants like Ambanis and others puts big capital into it..
True, but building the right kinds of models with the right weights seems to be the first hurdle since all companies who have made foundation models have used the vastly available free data from the internet.
Weird if polygon can be built for scaling etherium why not wrt gen ai??
Mainly because there is a huge resource crunch, especially at the hardware side of things. NVIDIA mentioned in their last quarterly earnings call that they are STILL behind shipments to their largest customers and yet they expect a large amount of increased orders in the next quarter. So even a united front within India may not be enough. Unless someone figures out optimisations that can have deliver compute on existing GPUs to train new models.
I think he meant "'talent" but also capital. America is entrepreneurial driven. The best for india is to compete a the national level instead of international.
You need tons of cash to keep AI up and running and data infratructure thus resolving scalingb problem at a global level.
Talent alone is not it. There are tons of Scandinavian kids that are also talented.
Money follows talent. In the chronology, research comes first, then capital follows. The important part is change in Indian mentality so it actually starts valuing research and PHD.
@@esycommerce it's like asking people to "change" and people hate "change".
@@esycommercetrue, also there are a lot of things apart from ai where the intent (research) is needed and as compared to others we lack that .. I won't comment if its a good thing or bad things .. but India has enough resources .. but no intend in this direction ..
@@h.c4898 Some more than others.
@@10101amit Absolutely! We keep talking about our age old sages and the theories in their scriptures. But how many of us have actually gone out to prove (or disprove) them? The entire concept of scientific approach rests on the questioning authority and relying on evidence. How much does the Indian culture support that thought process?
Why do people compare market cap to GDP of a country, there is no comparision on that bases because GDP is amount of goods produced and market cap is not amount of goods produced!
I know it's not a good comparison but it still gives a rough idea.
@@zenmaster4829 it dosen't, market cap is sometimes 30-40x the revenue
@@polaris1985 Bro literally all the major indian news channels have compred the market cap of Tata Group with Pakistan's GDP but everybody have problem with comparing Us tech giants market cap with India's GDP .
Better comparison will be to compare,
Profit of that company to National GDP.
As GDP is net total income of everyone.
@@Charvak-Atheist revenue would be better, profit is revenue less expense.
Bro, Really great content. Keep making such an amazing videos.
I've subscribed to your channel.
Thank you for the uplifting words. We're preparing our next video as you read this!
Great content. I hope you remeber this short comment when you reach a million subs!
Well, you're the @motivationmoghul. Consider us motivated! Thank you for the encouragement & be sure to subscribe.
thanks
You're welcome!
❤
Mean while I’m ChatGPTing how to beat OpenAi😂
Is that even legal? 😅
You need money for everything. True innovations and breakthroughs happen only in places where you have funding. You can't expect breakthroughs when more than half of the population is living in survival mode. Its not surprising most will chose tried and tested paths.
True. But as a society India needs to nurture a fundamental change within itself that values education and research with money as well as respect. Japan did it after World War 2.
@@esycommerce Too late for that imo. The A.I. Research wing in China started in 1980.
Russia has money and more money than India
It's crazy to give the impression that the tech is only India and Silicon valley.
How about south Korea, China, Japan, France, UK, Singapore...
@@mihirvd01 And that's exactly our point. We could have and we should have, but it doesn't look like we did.
well neuroflex framework is completing and nextgentorch is coming soon in-depth of research it is coming one man army
Just browsed over your article on blogs.kasinadhsarma.in/neuroflex-features-documentation Is this something you're working on?
saravam is finetuning foundational model
Noted
Indians never cant compete on this space.. i am IT Leader on a LLM project but i can say, we dont have that capability!
building a foundational model in now a days is dumb and we have all seen what happened with bloomberg's financial model, gpt-4 has beaten it very badly, and I think in this AI race the people who has more money will win the race, It will be only possible for common people when the resources will become cheap enough so that everybody can use and build
Excellent 🎉
Thank you Amit, please share the video with your LinkedIn and business network!
Gpt Ai models are too expensive. India can't afford it simply.
That seemed to be Sam’s opinion as well.
Brilliant video and you did an amazing job in pointing out the key weaknesses of India in AI.
I wish to see a similar video on
AI scene in china.
Keep up the great work 👍
This sounds like a great suggestion! Allow us some time to look into it. Be sure to subscribe and hit the bell icon to be notified of or next video!
@@esycommerce Sure❤️❤️
They are not made in india
But made by indians only!!!
Because Indians have to go outside for research, the Indian Government does not pay as well for research
₹35000 Is not enough for a PhD they should be paid more than 1 lakhs
Does it matter!? They're now american citizens contributing to the researches and innovation in America 😂 There's nothing for India to be proud of as such. With the funds that I get in India, I'll never regret moving to US. I spent 3 months researching in data science only to get my work stolen by my proff!? what a downright stupid country this is where education/research has no value.
It's a dejecting scene so to speak
It's not just the Government tbh, even the private sector can be very well be blamed for the same.
You seem to speak from experience. Would you be open to sharing your experience with us in detail?
Nice work
Thank you Ricky! Have you subscribed yet?
explain Mistral AI. the future will be owned by open source
We love Mistral too! What’s your most favourite thing about it?
@@esycommerce my favorite thing about mistral is how they attempt to fake benchmarks to make themselves look better while believing the public is stupid enough that they will get away with it. other than that their models are ok. I dont think they need to fake anything. nothing wrong with matching or slightly underperforming closed source.
I don’t get why India has to be in a race with the US to build a foundational LLM in the first place. Something smaller scale like Project Indus is a better bet. I’d rather that India focus on its development initiatives and allocating more of its budget towards improving educational facilities and encouraging more R&D in general rather than get involved in some “arms race” for false national pride.
That’s exactly what we’ve mentioned in the video. India needs to focus on strengthening its research capabilities and for that there needs to be cultural shift.
Yea we need real funding for innovative products ( but improving the current educational facilities will be the biggest waste, if we really try to understand how this AI /ASI revolution is going to change it all...
@@babyjvadakkan5300 I don't think the revolution will be as quick to make all investment in education redundant. Of course, I know we are developing humanoid robots but the phasing out of human jobs is not likely to be as uniform as we are seeing at the moment. The AI bubble will likely burst, and we will stick to the developments of narrow AI for the business use cases.
@@aaditkamat4995it's all about how u see this hype/bubble but are sure that we are going to stick with the same ways of development,if the the so called Artificial narrow inteligence can code , design, validate and plan ..(just accept that even though we will not ASI but GPT 5 or LLam 4 will replace most of our work by next year 😌
Unless India adopts CAPITALISM, it’s hopeless. Let’s face the fact. You need money.
Why do you feel India has not adopted capitalism?
@@esycommerce - What is India’s position in ease of doing private business ? How many acts and laws and forms are there to start a private business in India? What is the tax for doing private business in India ? How many major economic sources are still managed by government for example coal, banks, railways etc ? Why they are not privatized yet ? What is the general mentality of population towards Capitalism? Try to answer all these questions and you will get the answer whether India has already adopted capitalism or not.
A lot remains to be done. For all. Equally.
VISA , MASTERCARD UPI
Why not compare India with russia?
Why not indeed! Could you suggest a good starting point for us to research the topic?
@@esycommerce No. I mean why is russia not over-cited in this video like the USA?
Why are there no known Indians in russia? Why are they soooo few Indians in russia? I mean there are more Indians in Côte d'Ivoire a former French West African colony than in the entire russia? Before the war, there were more Indians in Ukraine than in russia. Today there are more Indians in Poland or Armenia etc than in russia. Incredible!.
@@francoislechanceux5818 Why do you feel that USA was over cited in this video given that virtually most ,if not all, notable Research, Development, Capital, Talent, Corporation and Paying Userbase are primarily from the US?
@@esycommerce That's secondary actually. The principal question is Indians in different countries. Why so little in russia? And why no successful Indian there in any field?
@@francoislechanceux5818 Looks like you have a point to make but want us to make it. Why not say your piece out loud? What do you want to tell us?
banias will never let indian govt. to invest in research because that will jeopardize their dhandas.
Why would ANY businessman NOT want to capitalise on innovation and a chance to have patent protected scalable products that gain MUCH higher margins than any trading businesses?
You are mixing up multiple topics into one. Creating a Foundation LLM and training it is one issue, while research on AI ( or any other subject) is another. Also, you are criticizing why there is no reusable spaceship while Bharat is still perfecting art of rocket for moon landing. Building requires time and consistent efforts and it is a series of linear steps. Europe and USA took 200 years, USA took 100 years, China took 40 years to achive current level of development while you want Bharat to achive same in couple of days 😂.
Research happens first and then comes training. We would’ve been able to train models had there been research to develop models in the first place.
The expectation is not for India to do something in a couple of days, but rather to work towards an eventual mindset shift that prioritises research and pays attention to it so that we don’t miss future innovations.
There was a time where China and India were at the same level, with India having a larger advantage in some cases. But that’s a topic for a different type of video.
@esycommerce you are again mixing up. Research is not unidimensional. From the AI perspective, one dimension of research is components, mathematics underneath components, architecture of systems, etc. The other dimension refers to developing products based on architecture and components. Another dimension refers to increasing the efficiency of the systems & product during training, deployment, and operation. There are more, but it is not a forum to write an article on AI research. Most of the contemporary AI products/ systems (barring a few frontier ones) are based on well established technologies. Any budding organization that wishes to develop a contemporary AI product/ system ( e.g. LLM) will utilize exiting available technology, not focusing on research. Once cash flow starts, research sprouts.
Also, to sprung research at country level, there is no one solution. There will be a multitude of initiatives such as three way marriage of industry, research labs, and universities ( like in USA), import of talent ( like USA did for its rocket/ missiles after ww2, China did across the industry in 1980s & 90s), glorification of people in sciences & mathematics, govt emphasis on primary science research ( not only on technology), acquisition of technology from other countries, emphasis on high tech manufacturing, etc.
So instead of looking into negatives, strive for solutions or at least do constructive criticism.
The Game that you talk about Playing in this video of yours is rigged against us from the Onset. You can no way compete with the Creator and Win the Game.
Every game is rigged to benefit some part of the society. Till someone comes in and changes the game entirely. UK was the global juggernaut till WW2 happened and USA jumped on the opportunity. You can either complain or you can Innovate. Choice is ours to make.
@@esycommerce You dont know what I am talking about
@@Ashish-no5ys Could you help us know more?
Very wrong question. These clowns like to live in a fantasy world where india is already a vishwaguru and superpower.
The first step to growth is the acceptance is one's current situation - the true situation. The more we ignore it with self pandering, the more we will keep lagging behind.
@@esycommerce Absolutely, acceptance of the current situation to pave a roadmap for the goals is a necessity for success especially at the global scale. After years of denial, Indian cricket took drastic steps that lead to sudden surplus of local talent who can perform almost immediately at the global events.
@@KhajaMD143 A great example indeed that drives the point home for us! Where do you think we lie in the 5 stages of grief - Denial, Anger, Depression, Bargaining, and Acceptance?
This guy is delusional. This is what NASA would say before ISRO began sending satellites. We don't need that kind of money to some the ai thing. Sure we are way behind in terms of skills sets like IT datacentres, software prowess etc as a nation but still sets don't solve problems by themselves. People and a willingness do.
Sachin could you elaborate your views? We’d definitely want to understand your perspective better.
The real fact is India dont have the foundation to grow big, even the most talented guy have to goto west to invent and prove his capabilities, in India he would be just another guy. You talking about AI, tell me one messaging app which works well which is made in India, forget messaging app, just name any app which the world is like "we cannot live without it", so how you saying Indians are the most talented software developers, may be they are but they want the support which they will never get in India, so India as of today will not even be able to spell "A" of AI.
I’m sure we can cover the “Artificial” part of the “Artificial Intelligence”. 😂
@@esycommerce Maybe, but I am sure 50 percent will not be able to spell artificial.
Highly misleading news. To build a reasonably intelligent system (A.I) you need two things - data and compute. ChatGPT was made by sucking in all the publicly available data on the internet and its still just a mediocre and dumb system. Please understand that we don't have enough data available in native languages to create even a dumb system. It simply doesn't exist. I am hoping that you already understand the second factor of non availability of compute. It also simply doesn't exist. Actually it also doesn't exist in the silicon Valley. That's why they are building it right now. Sam has quoted an amount of 5 trillion dollars worth of infrastructure to create AGI.
All that Indians can do is to download an open source Model like LAMA or Claude and fine-tune it to our local needs.
India is not even a participant in the A.I race. No point in ignoring the truth to support hollow jingoism.
Agreed with you on the point of how ChatGPT is made with basically most if not all publicly available data. It in fact took them some five years to get the first public version, I think.
The 5 Trillion seems more of a PR stunt and less of a need of actual compute. Yes, the compute needed right now does not exist either.
Just because India isn’t a participant right now does not mean it can’t participate in the future.
Jingoism is when one feels obligated for the results without putting in the effort and patience in building infrastructure and awarding real talent. We’re saying we need both of these things if we want have any chance to compete in the farther future.
@@esycommerce 5 trillion dollar statement isn't a PR stunt. We will soon witness trillion dollar training runs. Actually there's also a third requirement which is electricity. Private companies in America are leasing nuclear power plants just to train A.I. Before making any claims we should also consider a question. Even if we had the money, from where will u get the hardware. All semiconductor facilities have orders booked as far as three years in advance. There's a reason why America controls Taiwan. While Indians can certainly participate, India can never do it. Even if you some how miraculously scale up economy and get the infrastructure in the next decade, by that time Meta Open AI Deepmind would have multiple agentic AGI running all around the world.
@@esycommerce please note that Sam said that it's hopeless to try to compete in "Foundational Model" sector. Tech Mahindra is just playing a cheap PR Stunt. Neither is their model a Foundational one, nor they have developed it all internally. It's a cheap native wrapper on an open source model.
@@esycommerce we did such a similar pr stunt when India claimed to challenge Google on Geospatial navigation by releasing Bhuvan. By the way I have been working in the A.I field for the last 3 years with Amazon. I know what I am talking about.
@@naseebo8267 We'll try and contact their team to see if we can learn more. We're certainly intrigued to understand Indus well
Please invest in bathroom and police your streets, forget about Ai.
Ouch! Who hurt you? I mean, yes, we should invest in infrastructure and law and order. But it doesn't need to be at the cost of the next frontier, right?
India pehle road aur bridge bana le theek se
India needs to do both.
dont worry we have 4g motabhai and 'made in bharat' labels manufactured in ahmedabad. 🤡
Well, Ahmedabad, IS in Bharat, so...
Stop comparing and even asking.. it's disgusting that we ask such questions..
What is so disgusting about wanting to get better at the bleeding edge of technology?
Facts are fact
That they are!