Deepseek-R1 ‘Open Source’ Chinese AI Interview in English

I know sweet FA about all of this but there seems to be a lot of chatter about this “open source” AI from China performing at near levels achieved by OpenAI but at some fraction of the cost and computing power. Of course, it won’t talk about the Tiananmen massacre… I guess there may be all kinds of implications if it lives up to the hype.

There’s a translated interview with the CEO here:

2 Likes

Also a new podcast from China Talk

1 Like
1 Like

The true Turing test is whether AI can replicate human partisanship

There’s a transcript from the China Talk podcast here in case of interest. TL,DR: Deepseek abilities not surprise, China still behind US on AI, this shouldn’t be reason to abandon curbs on high-end chip sales to China, which is being affected by restrictions

The interviewee, an ex-OpenAI policy guy, has a 10:1 odds $20,000 bet with Gary Marcus that we’ll see AGI in 2027.

Lol (while crying for my pension assets). Not just “Nazis” melting down then. To be fair the futures are off lows, fwiw. (But, no, let’s not turn this website into a speculative meme chatroom).

There’s this interesting article: The Short Case for Nvidia Stock | YouTube Transcript Optimizer

There are many interesting points in it, including the shift of computational cost from huge amount to train, very cheap to use to… also rater expensive to use, but exceptionally good (one quoted example is the coming O3 model burning > 3k USD on a single task). Plus, the rapid advancement of hardware meaning that, in order to keep up, companies have to build larger and larger data centres with newer and newer tech, which basically become obsolete in a year or two.

1 Like

The guy interviewed in the podcast above is still bullish (cheaper ai, more ai, more chips), fwiw. But then he’s expecting AGI in 2 years, just left OpenAI and I guess is still vested.

1 Like

The Turing test will soon be irrelevant.

In a few years’ time, human (?) intelligence and artificial intelligence will be fed on the same substance, over 95% of which will be produced by AI, and will therefore be indistinguishable.

1 Like

It seems the AI race is the new space race. Only, rockets and spaceships didn’t take our jobs. I wonder what will happen once that starts happening in even larger numbers, more visible to those living in the West.

That may be premature (unless all your assets are Nvidia stocks :wink:). If AI/ML is a productivity booster (that’s a big if, but it may happen), then there could be a general increase in productivity and economic growth in the medium run, as the technology becomes cheaper to users. See John Cochrane on the topic here.

So far, all advances in technology that substitute capital for labor have lead to economic growth and the labor displaced by them was readily absorbed into other production sectors (starting with the cotton spinner). At this point there is no reason to believe that AI/ML is different. But it is still too early to judge its impact.

I am more concerned about AI/ML in warfare. It is happening now (UAV recognizing targets, etc), and seems to be a use case that is not talked about in proportion to its importance.

Yes, agree with this, I think, and thanks for the links. There’s a book I own but haven’t yet read, but the theme has been outlined elsewhere now, called Shock of the Old about how old tech sticks around. Famous example is how factories kept their existing drive shafts with electrification, just replacing the steam engine at the end of the line rather than electrifying each machine to allow for greater efficiencies. Don’t know if it’s exactly the same with digital assets. Some evangelists talk like whole social structures will be upturned, though presumably they don’t mean existing wealth structures as they’ve got the buttered side of the bread already. I don’t know if that’s why the sudden taste for disruptive forces, not just in industries and markets but other areas (being neutral here).

Edit: I wrote this, then thought what a load of rambling jibber and meant to delete it and somehow posted it instead…

This PR provides a big jump in speed for WASM by leveraging SIMD instructions for qX_K_q8_K and qX_0_q8_0 dot product functions.

Surprisingly, 99% of the code in this PR is written by DeekSeek-R1. The only thing I do is to develop tests and write prompts (with some trials and errors)
(ggml : x2 speed for WASM by optimizing SIMD by ngxson · Pull Request #11453 · ggerganov/llama.cpp · GitHub)

I guess that these scenarios generate more readers/views, while “ML may enhance total factor productivity in the medium run, let’s see” is relatively boring (unless you are an economist, or understand compounding, in which case it is super-exciting, as eg even 1% extra growth gives 50% more GDP in 40 years and has wide-ranging implications for pretty much everything).

Yes, getting generalist journalists to care about steady, incremental but compounding growth is quite hard when they think all news has to “break”

By chance, just read this intro:

I relatively regularly find myself arguing with well-meaning beneficiaries of economic growth since the Industrial Revolution who say we have to bring it all to a halt to save the planet. Anyway, I’m way off track…

1 Like

Sure, “we’re all gonna die”, “the economy is about to collapse” are very click-friendly titles (as long as they don’t get lost among a few hundred more…), but I do get the feeling that in the past few decades we unleashed change at an unprecedented rate, and a kind of change that not only affects “tooling”, but also our psyche: the internet, omnipresent mobile network, social media, and now AI. I think we are really going towards a future where there will be not enough work for everybody; less and less will be worth doing using humans.

Bored children are always ‘problematic’, ‘bad’ children, but I think adults are the same (only we call them ‘lacking purpose’, ‘depressed’ and so on). In theory, leaning back and just enjoying life sounds good, but I’m not sure everyone could find something to fill (give purpose to) their lives.

Also, companies need consumers, consumers need money to spend in order to consume. Basic universal income may be the answer, but the period of transition could be harsh. Once a few million people set cities on fire, the governments will have to do something, but I’m worried how that will go.

3 Likes

Yes, yes and yes, I think.

Given how most of the developed countries are not having enough babies to replace the aging population, we will soon have a exceedingly large number of old people requiring elderly care, so not all jobs are lost and a lot of new posts will open up :smiley: Unless Tesla sells an elderly care tesla bot. /s. The aging population really is a ticking time bomb and only time will tell what happens. South Korea will be the first big example of it, we should all learn from it and hopefully adapt in time…

1 Like

Not everyone is suited to work as caregiver.
The more a job needs ‘the human touch’, the less likely it is to become automated. But, again, we cannot ask become nurses, teachers, hairdressers and so on.

2 Likes