We're creating a new top-level product group at Meta focused on generative AI | Hacker News Reader

We're creating a new top-level product group at Meta focused on generative AI

facebook.com

114 pointsppsreejith129 comments

Has Mark become distracted from his VR obsession by AI?

It's so crazy but Mark's personal boredom or interest in either of these could swing the whole direction of computing one way or the other.

I do think VR could be the ultimate application of generative AI - one of the biggest bottlenecks at the moment is how expensive and tedious it is to make the 3d content necessary to fill out expansive VR worlds. It could also be central to actually making viable things like codec avatars which currently are stupendously compute intensive but perhaps don't need to be if you can generate the output using AI instead of actually coding and decoding it at the other end.

> VR obsession

Seems uncharitable. He's making a long term bet. Nothing wrong with that. People would complain if they never pursued any new avenues.

> I do think VR could be the ultimate application of generative AI

Then it seems a good thing for them to pursue then?

Yeah I agree - while I think Mark is showing some pretty knee jerk decision making lately (copying twitter subscription, chasing AI), I think the massive investment in VR shows a vision - if it will pay off or not is up for debate. And this is from someone who severely doesn't like him, there's plenty to good reasons to dislike him but I think the VR move is more strategic than people give it credit for.

> I think the massive investment in VR shows a vision - if it will pay off or not is up for debate.

The Softbank way!

Yeah, the key difference being softbank has a large enough portfolio it theoretically doesn't matter if most of them don't work out. I think it's a poor long term strategy, but that's with hindsight, and I understand the reasoning.

Captain Ahab made a long term bet, too.

As did Frederic Tudor. His bet was similarly unprofitable in its first few iterations, but now you can buy ice by the tonne in New Delhi.

I don’t mind when rich people make huge bets with money that’s within their control.

Facebook's investment on AI/ML has been huge and had a great contribution on the field. They are not new players chasing a fad. They foresaw this long ago, as did many other big companies too.

Here’s some recent ML work from Meta that I really like:

Toolformer: Language Models Can Teach Themselves to Use Tools

https://arxiv.org/abs/2302.04761

also this https://arxiv.org/pdf/2009.02252.pdf

They foresaw it and the best the CEO could come up with is "multi-modal experiences" and "more photo filters"?

> Has Mark become distracted from his VR obsession by AI?

Possibly more like "people in our VR spaces aren't finding many other people to engage with. Let's generate AI partners that are hard to tell from real people to make up for that."

Oh man, that's kinda terrifying. Then suddenly "everybody in the world" loves Facebook and shills them every chance they get

They'll call it 'seeding', and claim Reddit did the same thing a decade ago when it too was barren.

At least those seeds were by humans, and if you replied, you replied to a human

I think it would be beneficial to have even just ChatGPT level AI bots in virtual worlds. Getting a lot of players online in the same location is hard both network wise and socially. Bots could fill that gap especially for games. Empty VR worlds feel weird so this could bring them to life and free you from designing only smaller spaces or natural landscapes to compensate for the lack of people.

While synthetic worlds are still tedious to create, “nanite” tech in Unreal Engine 5 makes it dead easy to recreate real-world environments within a virtual world. Heck, you can buy almost real forests and cities - enough to create a full-fledged game - for under $200

Well with somewhat justified mockery for his metaverse vision. Meta's reality lab isn't doing too good. minus Quest/Quest2, the execution for XR is just really sloppy, even John Carmack can't stand the pace of Meta[1], fixes takes 6 month(!) to push through.

Even if Zuck wanted, his team doesn't seem to be able to pull that off, be that of tech limitation or just too many value-extract in the team. It doesn't seem XR is the reality. at least from forseeable futures.

[1]: https://twitter.com/ID_AA_Carmack/status/1626389610635923457

It's possible that he's noticed how much in the way of continuous press that ChatGPT and Bing have received, compared to the mocking that his VR Microsoft Office has.

I 'm glad i got the OQ2 at the initial price. It's a great product unrivaled for many years, i just hope they dont brick it.

He's talking about text and possibly video here, not VR.

Generative AI is key to passthrough AR that goes beyond simple greenscreen-like overlaying and does things like transform people and the environment being viewed.

So it seems innovation is now completely non-existent at FB, and their new modus operandi is to throw shitloads of cash at any new fad that surfaces in the media and see what sticks.

Thats tells you a lot about whats going on and what kind of people work inside there.

They seem to be a dead man walking.

A simple Google search would show how much FB has contributed to the AI/ML community.

I hate seeing these comments in HN. No substance. Just saying whatever to jump into the FB hate bus.

It's a valid point though.

There's a difference between running a research lab and a product organization. I used to work in a research lab (Nokia Research). Lots of cool research. But product divisions routinely ignored what we were doing. Not invented here, not a priority, we're busy moving deck chairs around on our Titanic, etc. Very little industrial research translates into customer/product/share holder value.

Facebook, like several other big companies, pays a lot to keep some very smart people on staff. They constantly come up with new stuff that they publish about and that also result in patents. That's not the issue.

However, from there to product features is where it goes wrong. Most of Facebook's product portfolio hasn't changed substantially for quite some time. And they haven't really launched any noteworthy new stuff either. Publishing a paper doesn't really count. It's nice and it will no doubt help other companies to succeed where Facebook is clearly struggling to succeed.

That is my outsider impression of what's going on at Google. The ChatGPT interface that Microsoft is rolling out absolutely should have come from Google first. I'm sure they actually have the technology, and they've probably had it for years now. What Google lacks is any sort of leadership or discipline to turn a prototype into a marketable product.

I didn't want to bring it up but, yes, Google is obviously struggling with the disconnect between what they are doing in terms of research and product as well. And they are pouring billions into the research. They even spun off alphabet as a company to do all the blue sky research. With so far not a stellar track record. Mostly their revenue is just Google running on autopilot and vacuuming up ad revenue wherever it can.

Most large companies struggle with this. A lot of former industrial research labs have either disappeared entirely or became a shadow of their former selves. IBM, HP, etc. used to do lots of blue sky research. Xerox Parc was a big deal and they obviously failed to productize. But of course inspired Apple and quite a few others. It's a tough topic. I don't think many companies have figured this out.

There are positive examples of course but mostly the return on investment for open ended research in a corporate setting is not as high as it should be considering the valid research that gets published and then fails to get monetized by the companies doing the research.

OpenAI, is the smart counter move. Spin off a company, fund it, and then license back the technology. MS seems to have learned from it's past efforts in MS Research.

While there have been a few public failures that were rather ... embarrassing ... and had their plug pulled off quickly, I don't think anyone can, in good conscience, claim that Meta has not been supportive of the AI/ML community, both in terms of research output and in terms of tooling [1].

[1] https://pytorch.org

Exactly. These are all R&D tools and papers. I think the criticism is quite valid if you look at actual products.

It’s also a fair criticism of most of big tech wrt AI, and a lot of AI/ML: there’s a lot of cool-sounding R&D, but so far there are few (arguably no) resulting products.

Maybe we’re starting to see that change, maybe not. Bing/ChatGPT is promising, but ... nascent. A lot of this has echoes of the self-driving revolution.

Whenever I read techno-optimists on this site, I cannot help but think back to 2014-ish when they all predicted that there would be no truckers in 5 years and dramatic changes to infrastructure to support a wave of self-driving vehicles.

It is also questionable EVEN IF Facebook is just throwing cash at fads and has stopped being innovative. That doesn't mean shareholders will lose out which is the ONLY thing that counts.

The inevitable nature of large corporations is to become bureaucratic and slow, yet many of them exist for decades without being innovative. It's sufficient to just buy out whoever is innovative in the market - Microsoft just bought into OpenAI big time. Other than incrementally developing its 40yo operating system Microsoft copies AWS, Netflix for games, Sublime Text and Macbooks while throwing some random cash at garage projects that amount to effectively nothing that can exist without subsidies (Hololens).

You are missing the "now" in the parent comment.

They are talking about actions in the last year or so.

Meta's open source projects that they own are maintained now by eager volunteers, not staff.

Anybody could see how Facebook is floundering, Instagram can see the iceberg approaching, and changing anything major in Whatsapp will piss off literally billions of people.

They've cornered themselves, so now they're reaching for literally anything that could save them. That seems reasonable as a failing company, except when every action they make affects entire markets and legitimizes trends on a very shaky basis.

Pearson Airport in Toronto is voted one of the worst airports in North America. They routinely lose your baggage, delay the flights indefinitely, people have to sleep on the floor, etc. - totally dysfunctional airport. This problem is not new - it has persisted for years. Finally, they found a solution: throw AI at it! Here's the announcement: https://www.narcity.com/how-pearson-airport-air-canada-new-a...

And now, this meta thing... :-)

Pearson is not the worst but depends on destination. (International flights have a nicer terminal, but national flight has been some years).

That said, Air Canada (also mentioned in the article) is also investing in AI while their default online booking process is still quite bad. Last time I wanted to book a flight with a pet, the website told me to book the flight without pet, then make a phone call to tell them I would be taking a pet with me (and cost would then depend on destination) just like two decades ago.

So I decided to use KLM and booked my flight, with pet, all through the website without needing a phone call.

The one and only time I’ve flown out of Pearson internationally to go back home, it was in the top three worst airport experiences I’ve ever had out of hundreds of international and domestic trips, and the other two were understandably due to weather, while this was a gorgeous day with no delays.

Security took me almost three hours to get through, and they had a byzantine system where once you got to the end of what you thought was the entire line, you’d turn a corner to find another long one, and that happened six times.

It basically devolved into a system where it was just a huge crowd standing around, and as peoples’ flights were about to board they’d get moved to the front.

It reminded me of the freakonomics episode where they talk about the LIFO queue: https://freakonomics.com/podcast/what-are-you-waiting-for/

I'm not sure how airport logistics work, but isn't it the airlines that lose your luggage, as opposed to the airport itself? If that's right (I'm not sure it is) then it would seem unfair to criticize the airport for implementing facial recognition for boarding your flight. They are two different things — there's the airline issues and then airport-wide issues, no? Happy to be corrected.

I don't know who is responsible for what, but in this particular instance, they work in tandem to create the most stressful experience for the passengers.

https://www.cbc.ca/player/play/2056928835691

https://www.gominplanet.com/blog/air-canada-the-worst-airlin...

My advice to anyone using Air Canada: upon arrival to the destination, don't waste time trying to find your suitcase on the carousel (it's not there anyway) - instead, go directly to the kiosk to complain about the lost luggage , otherwise you will have to stand in line for 2 hours. :-)

Yikes, this sounds awful

I always thought that baggage handling at the airport itself was done by the airport as one of the services it provides to airlines.

From the article "Pearson Airport's AI system is used during airport security to detect dangerous or prohibited items."

Uh, ok, I thought all airports have already been doing this for decades? One time I got pulled aside because the scanner thought my chocolate bar was a bomb.

"AI"

It's facial recognition.

their new modus operandi is to throw shitloads of cash at any new fad that surfaces in the media

Cryptocurrency. Voice assistants. VR.

What else?

You have to think Big, not just in terms of single product areas like VR but how it fits all together in the Metaverse. Zuck is a genius in this regard

I can't tell if this is parody or not, which admittedly makes the comment 10x better.

Did someone say AI on the blockchain... inside Metaverse?

Leveraging computer vision to transpose virtual dog ears on to your photos and videos? Seemed to work out for Snapchat, briefly.

Nope, they are the shoulders of a giant on which researchers and programmers are building the next AI tools and products.

See the open-source code repo of Meta Research. https://github.com/facebookresearch

Facebook invented PyTorch

When did facebook innovate? They are known for copying everything from everyone. FacebookAI is the most important think to come out of them , but it s not really related to FB's core business

PyTorch, React, GraphQL, Jest, FBOSS for data center networking, the list goes on.

Companies dream of creating just one of these.

Not to mention tons of contributions to VR. They may have mismanaged the PR on that, but the product, as it exists is quite a thrill.

Didn't they also contribute heavily to PHP?

Yes, their virtual machine was amazing. It brought all of the JIT technology and really sped up their PHP stack.

Now which of those are used by facebooks users

Well, let's start with React. It's used by 100% of Facebook users.

React is an innovation, in what exactly? It bloats the web with needless javascript and abstractions. It's a convenient shortcut for people making native mobile apps. It's not taking the world forward

Almost every request is graphql. ML also… all the ads and feeds use it.

rocksdb, zstd compression, buck build system are also noteworthy that have far reaching impact across the industry.

PyTorch was based on Torch, which facebook did not create.

I'm not fan of Facebook but what about React? Also if we are talking about "AI": https://mapwith.ai

People don't go to facebook because of react however. Sure any big company solves their scale problems in innovative ways.

You are the one who set up the goal posts. Just because you don’t like where you put them up doesn’t mean you can change them once people prove you wrong.

Nowhere did you say that Facebook’s innovations have to be the reason the end user uses the product, or even used by the end user at all. All you said is that they have never innovated. That is the bar. And it is clearly extremely low as Facebook has done lots of innovating with products used by developers all the time.

I know we are in a bubble. Facebook did not grow because of graphql or vr, it grew because it copied and bought anything in sight. I guess most ppl are too young to remember, however.

I guess you could say that fb created some innovations, mostly in their labs, but they are not in the business of innovation

> Facebook did not grow because of graphql or vr, it grew because it copied and bought anything in sight. I guess most ppl are too young to remember, however.

> but they are not in the business of innovation

Who in this thread claimed they are in the business of innovation or that something is innovative only if it affects their direct customers? This entire response is a strawman, plain and simple.

Once again, your clearly defined bar is, and I quote:

> When did facebook innovate?

Stick to the bar you set and be humble when it’s clearly proven to be a dismally low bar.

i did. i said their FB AI innovations are not related to their core business.

As much as we try to sugarcoat it, facebook is not known as a mobileUI or languageML company

Please. This is an escape hatch attempt to avoid being accountable for setting up goal-posts which have been easily met whilst you trying to also move them.

Even outside of AI with PyTorch, Facebook invented React and has been pervasively used by hundreds of millions of developers and billions on the web.

If neither of those are examples of Facebook's innovations on both the web and AI, then I don't know what is.

> on both the web and AI

any company with tons of money is going to build some stuff thats innovative by virtue of hiring people completely irrelevant to its core product. FB is a social networking and ads company and it was not particularly innovative at that. It was not a web development house or an AI powerhouse, and it's still not (that's not where it makes its billions).

> any company with tons of money is going to build some stuff thats innovative by virtue of hiring people completely irrelevant to its core product.

So by your own goalposts, they DID release innovations then?

> FB is a social networking and ads company and it was not particularly innovative at that.

I wonder what those psychologists and engineers who worked on the Like button with the team who built the AI that is used from that, to Ads and the social graph would say about the billions of dollars and stock value that it has made for Facebook.

Again, you set those exceeding low goalposts for your point. If not even that is an example of an innovation relevant to Facebook's business and made it billions of dollars, then perhaps nothing is.

Well it just so happens that most of the things people replied with (in fact, all of the things except PyTorch) have nothing to do with AI innovations, so great we can all agree to call them real innovations. Which is what they all are. And what you claimed Facebook doesn’t do.

The fact that you’re continuing to double down is astounding.

It's more astounding that people forget what facebook is, what facebook does, and who is its target audience (not developers).

Or people calling Oculus and FBAI etc as facebook innovations. They are acquisitions of existing technologies, paying the way to innovator status

Those would all be great arguments… had that been where the goal posts were set.

But, referring back to all my previous posts, you set them at “when did Facebook innovate”.

Maybe next time you comment, you actually read what is being criticized and stick to that, rather than this strawman you keep putting up.

I can’t insist that you’re wrong, but note that it is a very normal thing in business for the leader in any space to devote large amounts of investment cash to covering moves that might turn out to be disruptive.

IBM famously did this with the IBM PC. They didn’t end up holding the lead in computing, but they certainly stayed relevant for another decade while their mainframe and minicomputer competitors struggled to get into personal computing but were late to the party.

In the pop-business book “Bottom-Up Marketing,” this is given as one of the key strategies to being the leader: Play Defence.

I wouldn’t say that a company playing defence by throwing cash at every fad that comes along necessarily means they are dying or dead. It just means they know they have to have a finger in every pie, “just in case.” It’s insurance.

When was that not the case? Only time I can think of is A/B testing facebook.com and maybe the transition to mobile feed?

Their original idea came from theft. There was never any innovation, they just happened to get lucky.

What product innovation has Google produced in the last 5 years? Microsoft?

Reminds me of Zuck announcing he planned to make a personal assistant back in 2016 to compete with Siri and Alexa:

https://m.facebook.com/nt/screen/?params=%7B%22note_id%22%3A...

We're starting by pulling together a lot of the teams working on generative AI across the company into one group

"We will start by interrupting all of the initiatives currently in motion to introduce new layers of management and bureaucracy"

Maybe it will pay off.

Senior Management going to Senior Manage

I have a free idea for Facebook...

Let me fill my friends list with new AI friends who are like my actual friends, but cooler, kinder, and more helpful.

They can also introduce you to interesting news articles around the time of the next election too.

This will probably be the most likely use of AI. And where the money will go.

You'll no longer need an army of cheap labours in the third world to manipulate the online commons. Just a couple to get around capchas.

> They can also introduce you to interesting news articles around the time of the next election too.

I have family members who will do that anyway. Fox news even floated some interesting theories post-election.

Facebook sees that and will raise you digital resurrection of your dead family members to tell you they think [product] is amazing and really works!

and

You support [political candidate]? They raised you to be better than that!

that's what he s talking about, no?

I for one welcome the inevitable creative application of these AI "personas" and whatnot to be tuned for arguments and set against each other. A non-friendly, no-holds-barred debate. Full vocabulary and historical references allowed.

Then we can put it textbooks for study on creating tools that can be way too easily weaponized in a sprawling environment like the Web in current form. My money is on Poe's Law before Godwin, but by narrow odds. What, y'all thought this would end well? Go re-watch T2 damn.

Facebook building Shiri's Scissor for real? I'm sure that will end well. /s

For those who, like me, forgot what "Shiri's Scissor" was about: it was the "programmer uses neural networks, discovers wedge issues" story.

Shire’s Scissors are hypothetical statements optimized for maximum controversy, so effective as to be a potential cyberweapon.

https://slatestarcodex.com/2018/10/30/sort-by-controversial/

I will say this may be starting to cargo cult a little bit.

Do I think “generative AI” is huge - yes. But I also feel like people are starting to slap generative AI label on things somewhat motivated by silly reasons - marketing and pleasing investors who want to “keep up”. Or employees that want to work on something cool.

Well, the stock is up 3.24% today, so mission accomplished.

and is over 100% up since it went down to $88 three months ago and almost no-one here bought the stock.

They only bought into the media fear and nonsense that Meta was going to $0.

I'll withold judgement until I see CIO.com's survey of CIOs on this matter.

I expect $4bn in and it still won't do legs

Disclaimer: I noticed the tongue in cheek but wanted to comment anyway ;)

The legs discussion is way overblown and happened only because the "Metaverse by Facebook" has nothing interesting going for it.

If it did offer a product people wanted, floating avatars would be nothing more than a running joke that nobody would really care about.

But as it turns out the desire of the FB C-suite to own a software platform is not a particularly good selling point and here we are.

> A running joke

On purpose??

https://www.urbandictionary.com/define.php?term=has%20legs

$4bn and it doesn't have any legs.

probably should hire 10,000 more people, the AI race is heating up

I too love messing with the bot scrapers that decide stock trades.

Endgame for all of this, mark my words will be "immersive VR"

Why play hogwarts legacy when you could go to meta hogwarts and actually talk to harry potter about your day? People think parasocial relationships are toxic to society now?!?

HAH.

Just wait till we start seeing headlines about people killing themselves over their favorite character (or worse ai wife/husband) getting digital dementia due to dataloss or being taken offline forever. It rings back to that movie HER the one where loads of people fell in love with the same AI model, except in that movie the model had agency and wasnt insidiously designed by a major social media/ads company to drive traffic and profits.

I know my comments might come off as inflammatory, but I want to make a very clear point that I hardly have seen mentioned elsewhere. The problems that we see with the terminally online and other forms of unhealthy attachment to online personas will only be amplified as VR and AI get better.

> In the short term, we'll focus on building creative and expressive tools. Over the longer term, we'll focus on developing AI personas that can help people in a variety of ways. We're exploring experiences with text (like chat in WhatsApp and Messenger), with images (like creative Instagram filters and ad formats), and with video and multi-modal experiences.

I guess that makes sense. It works great with their metaverse vision (and might give it the USP it desperately needs), while bringing short-term benefits as well. I wonder what they mean by text in WhatsApp though? A chatbot for companies to paritally automate all the business that now happens over WhatsApp, or a text-mode Alexa, kind of like XiaoBing (which far predates the current GPT craze, but is now much cheaper to create)?

It's interesting to see the comments on his post. A good percentage is just "very impressive". What's the point of that? Just hoping he'll see your name and say "good employee"?

Probably some internal process monitors them, or they're trying to get noticed. I assume being top comment on a popular post gets lots of profile clicks.

You see similar activity on LinkedIn.

you think they're real? do you think i'm real?

One of the reasons I almost never visit Facebook anymore is I don't trust them with my personal data. And even after being fined $5 billion - the largest privacy fine by far in history - they never come clean and admit any wrongdoing.

I will stay far away from whatever Zuck accomplishes with generative AI.

https://www.ftc.gov/news-events/news/press-releases/2019/07/...

> Over the longer term, we'll focus on developing AI personas that can help people in a variety of ways.

It's easy to imagine a future in which Facebook have a paid monthly subscription to a friend, partner, therapist etc.

> with video and multi-modal experiences

And of course they'll try to slap some VR on top of it.

I don't think they are able to offer a product without renting out consumers to advertisers.

To exaggerate a little, I expect them to develop something like the Truman show where everything is sponsored and all "characters" try to sell you all the time.

Virtual friends as a service? This can trap billions in VR headsets.

friend(s) that's been specifically trained on you and your likes, no less

> subscription to a friend, partner, therapist etc

that might be pretty cool actually. I tried a couple of shrink bots pre-chatgpt, they weren't very good back then, but I think with enough effort it might be possible to get them to a mid-level human shrink in this latest iteration. FB owns tons of conversational data to train on, too!

Or a diplomat:

https://ai.facebook.com/research/cicero/diplomacy/

First time on facebook in years and I'm astounded at how dated the UI feels.

Yep, first .facebook.com link I opened in years and my default browser zoom setting for that domain appeared to be 200%, so I immediately noticed that all the icons and logos are raster graphics (instead of vector like they're supposed to) and the quality is extra crappy due to poor resolution / excessive compression. Amazing to think that this is the "F" in FAANG.

Desktop Facebook is hilariously bad. It's as if they assumed that anyone on a widescreen viewport must be on a tablet, and blew up every UI element to a size that they consider clickable on a 7" Fire Tablet.

If you want to visit 2007-era Facebook, check out mbasic.facebook.com. It's the only way to access Facebook Messenger on mobile; the main mobile site redirects you to a download page for the Messenger app.

I am bemused that the approach here is to pull teams out of wherever they are closest to the products receiving these features and into their own top level. Zuckerberg enumerates what the objectives are (bots, assistants, image filters, etc.) but I would expect this change to open up gulfs in communication holding back shipping something of value.

It'd be one thing if there was a single umbrella Facebook product, but Meta is going the opposite direction.

Interesting to see the somewhat muted announcement from LeCun who has been (one might say, for good reasons) sceptical of “autoregressive LLMs” and the path to AGI requiring something quite different

https://m.facebook.com/story.php?story_fbid=pfbid02zHwANqWrZ...

It still stuns me that so-called ML experts have turned into the "just add more layers" meme.

I denounce LeCun for his political tirades and just generally bad takes, but this statement of his is a healthy skepticism.

The more he opens his mouth, the less I respect him. Yes I know he's the most cited scholar in NLP. It doesn't mean he isn't sometimes wrong.

As Steve Jobs said, you have to work backwards from the user experience and see how the technology fits.

Let’s see if they actually do that or just build a gimmick

Jobs also said he'd have to whittle his fingers down to use the small Samsung tablet, knowing that Apple was working on the iPad Mini. He would say anything if he thought it would help shift more units.

Based on their own press release, sounds like they failed that already.

Wondering if all this AR/VR stuff eventually becomes like the Voice semi failures seen by Google and Amazon.

Its more of an annoyance to have a voice assistant at odds with you trying to push advertisement and listening in on your conversations to sell to you better. Factor in that voice doesn't always get it right it becomes a nuisance.

Snapchat just added their own AI chatbot too. interesting to see major industry shifts over LLMs

At my FAAMG company, the new way to get funding for anything is to sprinkle some ChatGPT on it. Two years ago it was Kubernetes. Now it's ChatGPT. Just hilarious to watch execs fall for this stuff over and over again.

Anyone have a beat on what Yann LeCun's take on this is?

I imagine investing in Nvidia and possibly a company like Google who make their own AI accelerators (TPUs is it?) is a good idea? Is there going to be enough fab space for the coming AI revolution? With Apple taking all of TSMC 3nm output are they planning their own Mac Pro with external GPUs and a possible move to the datacenter too? It’s certainly an interesting time to be alive…

Google has low AI industry exposure. You’d be better off just investing directly in NVidia and AMD.

Productizing FAIR output, or is that too reductive?

“Oh good” said no one, ever.

Did he even write this or did some marketing PR team do it for him?

> Over the longer term, we'll focus on developing AI personas that can help people in a variety of ways.

So Facebook are getting into the SockPuppet persona business, has the US military [1,2] and other 5eyes+X countries put out a tender for new sockpuppets, because the existing one's are getting too easily detected?

Arent they causing enough trouble [3] psychologically manipulating people, causing Hurt, Embarrassment and Harm (HEH)?

Anyway those bots are still making money for some![4]

[1] https://www.theguardian.com/technology/2011/mar/17/us-spy-op...

[2] https://en.wikipedia.org/wiki/Operation_Earnest_Voice

[3] https://www.reddit.com/r/TikTokCringe/comments/vkv8hu/aww_th...

[4] https://markets.businessinsider.com/news/stocks/reddit-reviv...

Probably ChatGPT can produce more human documents than Mr Zuckerberg.

> in a variety of ways

is what what people write when they cannot think of a concrete example, or when they cannot get my boss to sign off on something...

It might be the alternative uses test [1] aka the brick test. I had this at 1st year secondary school. Naturally I was top of the class on this exercise.

[1] https://www.creativehuddle.co.uk/post/the-alternative-uses-t...

Ah, thanks, that is insightful. Not to hate on the brick test, that is a useful exercise. But I would have been on top of that class, too, I was an excellent bullshitter in school :D

Reply on news.ycombinator.com