How will US tech firms react to DeepSeek? | Inside Story

how will US tech firms react to deep seek the Chinese artificial intelligence startup says it can match Google and chat GPT at a fraction of the cost Donald Trump says it's a wakeup call for developers so who will dominate in This Bitter digital rivalry this is Inside Story [Music] [Music] hello and welcome to the program I'm Elizabeth panum investors have poured billions of dollars into artificial intelligence research and development over the past year hoping to capitalize on this rapidly advancing technology generative AI could automate countless tasks and change how many sectors do business the US

has largely LED this revolution but now a Chinese rival has emerged deep seeks models are faster smaller and a lot cheaper will investors still be willing to put up billions if a more coste effective alternative exists and who's in the best position to benefit from ai's vast potential we'll go to our panel in a few moments but first Fenton Monahan has this report deep seek has sent shock waves through the global technology sector the Chinese generative AI startup says say it can get results comparable to that of its us Rivals and deliver them at a

fraction of the cost and processing power this is going to be much more accessible especially financially to a broader audience and it really abends the plans for monetization strategies that a lot of the you know Western uh AI firms that have been working on developing the Deep seek team says it only cost around $6 million to tr artificial intelligence model compare that to open AI the creator of chat GPT which spends around $3 billion a year on AI training America has seen an AI investment boom with venture capitalist investing more than $60 billion in startups

in the past year on top of that Tech firms like Amazon Microsoft alphabet and meta are each estimated to have spent between 30 and $60 billion in development but analysts estimate that returns are still dwarfed by expenditures the Trump Administration has made the sector a priority announcing a half trillion dollar private sector fund for AI infrastructure the US President says the emergence of Chinese Rivals should push American companies to innovate even more the release of deep seek AI from a Chinese company should be a wakeup call for our industries that we need to be laser

focused on competing to win because we have the greatest scientists in the world even Chinese leadership told me that Tech stalks have enjoyed a good run for months but the release of deep seek sent those same stocks plummeting earlier this week with chipmaker Nvidia losing more than 177% before rebounding that's the real I think disruption for the markets today is not just cheaper but a heck of a lot cheaper 96% cheaper than what we're finding from the current spend in alternate models like chat GPT and Gemini consultancy firms such as PWC have predicted that AI

would add trillions to the global economy in the coming years with the us the biggest beneficiary but the entry of a lowcost Chinese competitor has raised questions about whether the sector warrants such major investment and whether the us or China will gain the most from the AI Revolution Vincent monan Al jazer for Inside Story let's bring in our guests Ray Wang is the CEO and principal Analyst at constellation research that's a technology research and advisory firm and he joins us from the heart of Silicon Valley Cupertino in California from Sydney we have Toby Walsh a

professor of AI at the University of New South Wales and author of Faking It AI in a human world and in Hong Kong we have Brian Wong an independent geopolitical strategist and fellow at the center on contemporary China and the world very warm welcome to all of you Ray I'll start with you in California it has sent shock waves where you are in Silicon Valley and in Wall Street how has deep seek been able to do this Elizabeth thank you um what's been going on around is uh we've actually been dissecting this and here in

Silicon Valley it's becoming clear that deep seek may have used the open AI gp4 model to train its model using a technique called distillation it shows that deep seek is basically rein verse engineering proprietary models and then adding training but without permission and licenses so we've seen this going on and we actually think that what they've been able to do is show that there is ways to take smaller models and smaller parameters and do it at cheaper cost which is really much needed right now because in the internet Revolution things were decentralized open many players

and cheaper unfortunately at this moment AI is closed it's centralized there's few players and it's very expensive and this is an important piece because right now we thought AI was going to take a lot of money a lot of data and a lot of power to be able to solve the problems and now we know that that's not necessarily the case Toby there are some that are skeptical you know the CEO of of scale AI said that he understands that deep seek has something like 50,000 Leading Edge chips that they can't talk about because of

the export restrictions that are in place by the US from what you can tell so far about how deep seek has done this is it truly disruptive or perhaps a little deceptive it's uh it's plausible that they've done it with as little resources as they're claiming I mean maybe they've used it slightly more resources to make it slightly better but just the fact that they've matched the same uh performance itself is impressive people were saying that China was six months or more behind the US it turns out they're not they're they're right up there with

The Very Best of the West so it is impressive um and it's not without precedent that we've see people take models um and really reduce the amount of compute needed to do them we saw that with ch GPT a year after CH GPT come out um some Works some researchers at the University of California Berkeley built a chat GPD clone didn't cost the $30 million that chat GPT cost it costs $300 so we've seen a a constant push to try and do it with more with less and so I'm I'm actually convinced that they've done

it with perhaps as little of resources as they claim yeah and Toby you're not the only one Brian regardless of exactly how they have been able to achieve this how much of a boost is this not just for the time Chinese Tech sector but also for the Chinese government which has been trying to build this Tech sector independent of the West I think the successes of deep seek are Testament to the victory of Open Source AI so rather than characterizing this as a victory by any single government or firms within a particular National economy but

it's imperative that we look at open source AI as a means of emancipating and democratizing AI development and ensuring that ultimately even the small players that Davids in a yard could also stand a chance against the Goliath now of course it's a temptation to romanticize the victory of deep seek and to attribute this to National models a China way I think that's um in Fr in in Frank opinion on my part I don't think that's necessarily warranted you know for the simple fact that deep seek can fact flourish precisely because of how horizontal how anti-

hierarchical and how fundamentally liberated and also free its um constituent workers designers and Tech talans were able to you know work cohesively as opposed to working for very strictly defined hierarchy and arriving at a common set of goal and that is to arrive at some semblance of breakthroughs in substantially reducing costs whilst also accomplishing the same level of achievement as open AI when it comes to of course you know the models that developed so I don't see this as a as a victory by any particular State and I think it's quite perilous to portray this

race as one that is zero sum this AI race to AGI for instance of Singularity as one that must be won or lost by Nations that is not a conducive narrative that allows us to really take stock of the actual risks on the way of accelerating and and you know going completely breakless in our Tech uh Discovery process it seems to be the narrative that's come out of uh the US administrations various us administrations for a few years now though and that is why they have been putting all of these export controls on China to

limit the progress and I want to talk about you know some of the losers from what's happened so far one of the biggest of course has been chip maker Nvidia which was the most profitable company in the world Ray you know deep seeks claim that it built uh that it was built at a fraction of the cost of industry-leading models because it used fewer Advanced chips caused Nvidia stocks to lose shed almost 600 billion dollars of market value on Monday that's the biggest loss in US Stock Market history we've mentioned that it's rebounded but what

could this mean for the US chip sector and investments in it well I think that's a short-term blip and anybody who knows how markets work is that this was basically an announcement that had psychological implications because everybody had to go back and look at revaluations right on whether US tech Giants and energy companies and real estate valuations of data centers was too high uh and we'll find out in the next few weeks whether it was too high or not uh but you will see that the Stocker Market will rebr and nvidia's stock will come back

but the real question isn't that the real question is what you're asking and what Brian was asking as well is can we deliver AI in a decentralized abundant manner as opposed to centralized scarcity which is where we are at the moment and how we actually do that is going to be very very important uh and so for NVIDIA they're going to sell more chips the question is whether they monetize those chips as quickly or not um and whether you know having that chip dominance makes a difference right because necessity is the mother of in invent

Innovation invention right China proved you can actually try to achieve some of these results without with less chips less power and more importantly you know less resources and and that's what people want to know is can we get the faster better cheaper in AI because so far the race to AI has been very very expensive but the reality is what they've done is they've done that but they've also attacked the US Stock Market at the same time so it's a perception of valuations and so it looks like a psychological operations uh because if you look

at some of the reports we saw a huge rise in chinese-based social media activ and in Pakistan and fake accounts in Australia and it seems like a coordinated attack on the US during Tech earning season on the other hand it's a reality check to us companies why are you spending $80 billion on data center development for companies like Microsoft Amazon and Google and that that also is an important piece absolutely like Donald Trump said I mean he called it a wakeup call Toby given everything that Ry has said how do you see the various sanctions

export controls that the US administrations have put on place have they been effective at all in curbing Chinese progress in the tech sector or have they in fact been counterproductive well the evidence is that they're probably being counterproductive they've actually forced deep seek um to be more Innovative to to not go the route that the US tech companies have G which is scaling which is throw more money throw more compute throw more gpus are the problem that's the lazy way actually get to get to intelligence uh but it's the way that you know most of

the tech companies in the US have been going about it um it's refreshing to see you know you can actually go the other way to try and be Innovative um and do it less I mean of course we know ultimately we're going to succeed uh with much less compute we know that you can do it with just 20 watts of power that's what you know the human head uses uh and that's what the ultimately you want it to run on your smartphone so the ultimate Target is to do it with much less compute um and

and so actually I think it's what it's demonstrated is the US hasn't won it hasn't demonstrated that China's you know going to win this AI race it's demonstrated that the US hasn't won and I think actually that explains part part of the falls in the US Stock Market which is that they had been baking in the fact that that those companies were going to be the winners it's not clear that those companies are going to be the winners anymore no and so early in the AI race too Brian what do you make of the timing

it's also very very early in the second Trump Administration and all of this happened in the first week is is it a coincidence or strategic perhaps well you know there's a saying right that sometimes in history there are decades when nothing happens and then there days when all of history occurs and I feel like we're sort of witnessing this as we speak but uh added aside you know ultimately I I do think that attempts to curb China's rise logically could very well end up backfiring on the states and that's for a number of reasons the

first is to note that frankly Mo's law is by no means the only determinant of the pace and progress and in terms of AI development as also Kumi law when it comes to energy consumption and the fact that chip packaging is going to become progressively more important as time progresses as we approach the inevitable conversion point of one nanometer and beyond that it really doesn't make any sense to have smaller and smaller chips of to pursue you know size for the sake of size his size is not going to be the rate limiting step secondly

the US should also be wary of turning away and deterring talents ethnic Chinese scientists stem talents and recruits bearing in mind the fact that many of deep seeks Engineers are f should never actually set foot upon the US who are homegrown talents locally and natively trained in China these are signs and in fact cutting off ties and also engendering this atmosphere of paranoia in the states for ethnic Chinese scientist could very well backfire on Us's attempt to be great or Donald Trump's pledge to make America technologically great again and finally and just very quickly you

know at the end of the day the world is large in both China and the US and at small and medium States as powerless as it may seem to be superficially are in fact increasingly likely to not want to take sites and not comply strictly speaking with a sort of semiconductor restrictions and other forms of advanced export controls that we see the US adopt and we see that from Southeast Asia in Malaysia in Indonesia but also increasingly from non-aligned third parties and corporations who frankly view that these restrictions on their behaviors are not only detrimental

to the corporate interest but also to the PACE and progress of technological Discovery so in short I don't see you written about you know what this could mean for Access for for countries that don't have a lot of money to spend on this technology to build these systems from scratch what that can mean for democratizing AI and I I do want to ask you about that in just a moment but since we are talking about how the impact that this has had on the US Ray where you are how do you expect Donald Trump to

react to what's happened well it's very hard to predict what Donald Trump is going to do but I would say that in general uh we we have been in a non-kinetic war with China for the past 10 years and I think this is what people are realizing and because it's a non-kinetic war it's hard to see it's cyber attacks it's uh you know Pig you know Pig fisting operations it's whole bunch of things that are happening right now and the challenge is really trying to figure out what is that right balance of trade right for

example you take the Tik Tock ban and the reason there's a Tik Tok ban is because we can't have Google and Facebook in China right it's a reciprocity issue more than a privacy issue right and so there's this notion of what are the trade Norms that are going on and of course how do we choose to engage and and right now I think you know what we see in the AI race it's about having energy lowcost energy it's having lots of data having chips and having the resources to be able to pull this together uh

and you're right in you know Sovereign Sovereign AI is going to be important because countries that can't afford it are going to need to pull together to be able to get those capabilities to the residents otherwise we have an AI divide uh but what the US will be doing is mostly spending time trying to beef up its capabilities you know whether it's attract the right science attract the right Investments and of course you know continue with open and transparent systems so that organizations can actually build and expand on the AI Revolution uh however this is

the piece that's important the US tech giant are in danger for one reason is there are slaves to the stock price and because they're slaves with the stock price they're thinking very short term and prices continue to go up instead of coming down in the internet age and this is where we're going to need a disruption whether it's coming from China or any country to drive down the cost structure and Brian do you see the us perhaps tightening export controls and Donald Trump has put together a number of China Hawks in his administration including Secretary

of State Marco Rubio does that give us an indication of which way this Administration might go it's um a very good question I would say that Trump and indeed the American political establishment at large of which Trump may not necessarily count as a part have developed this overarching view that to inhibit China's rise in is integral to America's geopolitical objectives with that being said I do think that Trump is more prone and perhaps inclined to wield um export controls as a bargaining chip as a means of getting China to come to agreement on other issues

and America views to be of importance including the trade imbalance uh as perceived by the DC establishment but also on territorial questions or on geopolitical uh agenda items at large in short there's more fluidity and flexibility there but just very quickly on your point of about the Hawks I would actually say that despite the hawkish credentials and proclivities of the likes of Rubio Walts and many others that Trump has appointed at the end of the day this is a cabinet uh comprising one half of very skilled and professionally you know veteran and seasoned folks and

one half of uh rather inexperienced and dare I say green folks who are united in their common difference and sycophantic support of Donald J Trump so whatever the president sees to be integral to the country or precisely his presidency's interest they will execute and some of them would execute very well and flawlessly that is to say if Trump is bent on securing a so-called Grand bargain of sorts with China in preserving a grand peace a cold peace nonetheless now I'd imagine that many of these folks despite their own personal views would indeed defer to the

leader in the White House and as such adopt a more conservative posture especially when it comes to Military and security deployment and also presence in East Asia so less of a an emphasis upon that and we've seen that sort with a recent slew of Pentagon picks in terms of the D Deputy assistant secret of defense and also other of technocrats appointed senior middle or upper middle position so to speak in a hierarchy on the other hand you'd also see perhaps more emphasis upon Shoring up domestic manufacturing and also domestic Tech developments in America including in

the eyes of trump that means giving a free reign and cart blanch to many of the leading tech companies currently driving forward AI research including but not limited to open AI although that of course would also turn on a relationship Trump and outman and that's another story for another day yeah I'm glad you've brought it back to open AI because as you yourself said the race it has been between America and China but this is of course about more than these two countries Toby what does this mean for the democratization of AI and for access

to it these developments I think it's very good news I I I think it's you bad news if you had US Stocks the last couple of days although you know they'll probably bounce back but absolutely brilliant news for the rest of the world really which is that not only is the AI race to be won but the AI is going to be easy to be democratized as Brian was was quick to point out very importantly what is most significant about this model is not it was very cheap um not that it was need but that

but that it was open as well you can download it I have downloaded it myself for free and and build upon it so you know it opens the promise that you know countries around the world are going to be able to build their own models and let's not forget you know the other key component um that we've we've touched upon is the data um you know uh deep seek where it seems we're able to profit from open AI data we're be going to be able to profit all of us from Deep seeks data and and

be able to you know build specialized models using our own high quality data so I I think this is you know for the planet really good news AI is going to be a force for good for all of of us not just for the tech billionaires in the US yeah on that point Rey something that we haven't touched on is of course like many other Chinese AI models deep seek is trained to avoid politically sensitive questions how much of a challenge is Chinese government censorship do you think it is a challenge for its aspirations internationally

well if you're offering it for free and it's an open system I think that's part of the bargain right you're getting it free but you're also getting the Chinese view of the world right and I think that's by Design so I think if users using that have to take that into account you know but what Toby was saying is true right if you can download that on your own own machine all you need are four CPUs or GPU it's actually pretty light uh you can do that people are going to work on stripping out uh

the Chinese censorship as well so there'll be models that'll be able to take that out and put that back into use um but what we are defining is really the fact that you know we need we really need to democratize and decentralize this but you also have to understand who's providing you that technology right if it's coming from the US so have the US bias if coming from China of the China bias uh and you're going to see that with other country specific types of models that are out there um every country is going to

want to put their cultural constraint there as well as responsible AI constraints because everybody has a different view of this and so we're almost at this point where when we were in biotech and we're doing biotech research like what kind of genetic research is okay what is humanity okay with and then we're going to see other bodies come into place trying to regulate where AI is at that point in time but right now it's it's the Canary Islands of AI it's freefor all uh especially in terms of trying to figure out how to get to

the advancements as quickly as possible and then of course Society typically reels in all those different areas through regulation over time so we're going to see that uh and we'll see that curve just like we did with the internet and Toby is that freefor all good for the technology right now well I mean a fundamental question is is comes back also to this open source which is that if everyone could download the models um any restrictions that you might put in whether it be about talking about tanaman square or bioweapons can easily be removed so

it is going to be more interesting more difficult more challenging to regulate um if you know open source wins as it open source has many many advantages but there also some challenges it throws up um you the internet runs on open source software it's been fantastic for building the internet um it looks you know there's a good chance that open source will win the AI race as well but then that does pose those fundable questions which is how do we ensure that it is used in responsible ways and Brian we're coming to the end of

our show I'd like to get to a a final question why is this race the AI race so important to both countries what is it that's at stake from the Chinese side there's an overarching judgment from amongst its leaders that the reason why the Soviet Union collapsed and fundamentally lost a cold war was because it failed to unleash um the sort of new productive forces uh of immense productivity and also harness the Holy Grail of technology and that is why from the point of view of the CPC uh winning this race even though it's not

a war in their eyes and they don't want there to be a war between the two parties is of Paramount importance and from the states of course I think Washington is trying to still trying to Grapple with the implications of a rising power in China one that is far from being able to displace America by the way in so many ways and yet you know the fact that China has accomplished so much over the past four decades has clearly unnerved many in DC is a concern that if the Chinese are indeed standing level to level

shorter to Shorter with the Americans that a world order might be one that is unfamiliar to those who had enjoyed the dividend of the post Cold War uni moment so the way I see it as a philosopher or come geopolitical strategist as they often think that both Powers would benefit from recognizing then the face of the risks of unbridled AGI and also unregulated AI there is no winner from a geopolitical point of view we all lose and yet will the powers that be come to realize that well I can only stay cautiously optimistic thank you

uh Ray are you cautiously optimistic oh I'm cautiously optimistic in general I think we're going to have a good arms race on AI I think what's at stake here is really geopolitical influence uh China is able to influence a lot of countries uh with their money with their technology and with their ability to uh develop those Nations the US has not been paying attention to the geopolitical space uh we've been asleep at the wheel for the last four to8 years uh and and I think we're paying the price for that and so this is a

war in all fronts kinetic digital uh and of course you're going to see more as we get into space as we get into the polar caps so this is just one part of the overall us uh International strategy uh and I think you're going to see a much more activist president than we have in the past and just lastly Toby because Ray mentioned space there are so many that are calling this a possible Sputnik moment you know referring to the Soviet Union's uh successful launch of an artificial satellite orbiting Earth I believe in 1957 how

do you see this event in the bigger AI race I I think it's an important stepping stone for you know what is going to be the next Industrial Revolution the last Industrial Revolution gave us in the west high quality lives doubled life expectancy um I'm optimistic in the long term that um we'll do the same for you know many more people on the planet although I must to say I'm a bit pessimistic in the short term it's going to be a pretty bumpy ride the next five or 10 years gentlemen it has been great to

get all of your expertise on this very interesting topic we really appreciate your time that is Ray Wang in Cupertino California uh Toby Walsh in Sydney and Brian Wong in Hong Kong and thank you too for watching you can see the program again anytime by visiting our website al.com for further discussion do go to our Facebook page that's facebook.com/ AJ insidestory you can also join the conversation on X our handle is AJ inside story from me Elizabeth puranam and the whole team here bye for now [Music]