Chinese artificial intelligence startup DeepSeek’s latest AI model sparked a $1 trillion rout in US and European technology stocks, as investors questioned bloated valuations for some of America’s biggest companies.
DeepSeek’s latest AI model rose to the top of the Apple’s appstore charts over the weekend, presenting a visible challenge to costlier models like OpenAI and raising questions over the hundreds of billions in planned spending on the technology by the likes of Microsoft Corp., Meta Technologies Inc. and Alphabet Inc.
DeepSeek buzz puts tech stocks on track for $1.2 trillion drop - The Economic Times https://search.app/1Hm1eJsGpGC271wu7
Genuine Tech question.
If the fact that everybody is downloading the Chinese AI app Deep Seek is causing the AI stocks to crash, should we all be downloading Deep Seek right now?
Chinese AI start up Deep Seek has really spooked the markets.
S&P 500 tech sector is down 5.8% today, and on track for its biggest decline since, well, since the last time Trump was president.
If a $6 million start up is enough to wipe over a Trillion Dollars off your Tech shares, your Tech shares are weak as.
What's that? AI turned out to be an over-inflated hype bubble?
Well, I'm shocked, I tell you. Next you'll be telling me that Crypto is all just a scam and that SpaceX is nothing but a satellite company with a clever 'We're conquering Mars!' marketing gimmick....
And if you think the markets are spooked by Chinese start up Deep Seek's AI app, wait until they hear about their rocket ships....
Deep Seek AI - Gets things a bit wrong for way cheaper than all the other models gets things a bit wrong for.
Deep Seek doesn't recognise the events in Tiannenman Square.
ChatGPT doesn't recognise Elon Musk's Nazi salute.
Pick your own level of wrong.
@ProjectFearlessness wouldn't it destroy the environment in China?
@jonn That's the advantage Deep Seek has. It uses way less power than the big tech AI has, so it's carbon footprint is tiny compared to them.
@ProjectFearlessness surely that can't be right. Surely still one query is like hundreds of google queries.
Did anyone actually verify this?
@jonn
I'm not that techy, but as I understand it the whole way DeepSeek works is based on having a limited number of processors and making them work smarter, using a lot less power.
DeepSeek shared a lot of info about it last week and it seems to be true, but I would recommend finding a good techy person to explain it to you. I'm not techy enough.
@ProjectFearlessness I was polite! 🙂
I'm reading deep seek papers and RL-Zero uses more electricity during training for sure. Reading on.
I think that you really shouldn't use the greenwashing argument. I bet they have some bullshit benchmark that shows that trained models are 10% more energy efficient or whatever. But that's bullshit.
@jonn
At the moment, all I'm saying is what Deep Seek claims, which is that it uses about 10% of the processors the Big Tech models use, which is a lot less energy. That's a great thing, IF it's true.
They might be lying, but we will find out it they are lying as soon as it's replicated. And if they are lying, that's the end of their business. So why would they lie?
@ProjectFearlessness anyway, show me the sources. I'm like half-way through the RL paper and so far all that I see is that their training is clearly less energy efficient. It would be funny if they actually claim 10%, because it really is within the bullshiting parameters.
@jonn
I can see you've made your mind up. I will wait and see if the claims are true or not. Thanks
@ProjectFearlessness it seems like you can't provide a source, which is suspicious! Please enjoy ¥20.
@jonn
Keep resisting. AI is a nonsense. I'm rooting for DeepSeek to come out on top because it won't destroy the environment like all the other models, but I have no intention of using any of them.