this is a Good Question, and the answer has a lot to do with the economics of venture capital. VCs need there to be a transformative breakthrough tech that offers massive returns bc their model is, to some extent, predicated on having a Facebook every once and while to offset failed investments.
I remember when Google Translate came out, everyone just understood like "this is neat for periodic personal use but it's often wrong and will never replace the average fluent human" and it was used as intended and is a great invention. Why did everyone forget this with the LLMs masquerading as AI?
this is why they were pouring money into crypto and the metaverse, this is why they’re doing the same with AI. it’s been around 7+ years without any large scale changes in consumer tech, and about 2 since interest rates made big IPO exits less attractive. they’re getting desperate.
before SVB went under, they hosted an L&L seminar on the fundraising environment at my coworking space and the takeaways were:
(a) valuation ramps are dire: median series B valuations were lower than median series A valuations for the most recent quarter they had data (this is really fucked)
(b) VCs are sitting on record amounts of cash and are scared to deploy it because everything is failing, but if they don't deploy it their LPs will start asking for it back.
They were optimistic VCs would start deploying in a quarter or two, and then 2023 was even worse than 2022....
also really highlights what a blow to the ecosystem the SVB bank run was....there is no equivalent entity in the market today handling enough sides of enough deals to have access to this kind of data (and be willing to share it with clients...)
FOMO mostly? VCs don’t walk on water but there is the perception that they are the only ones who have access to talent and can set up the next Google or Facebook. Those made so much money and it’s hard for people to forget that.
I think it's 50/50 "need for a unicorn" and "me who don't understand the tech throwing money at deals with cons in the background"
Altman using Open AI to fund his fusion company for example.
it's mind blowing when you think about it
the VC's know almost nothing about the things they're funding
most CEO's are not experts in the relevant fields either, so they hire ppl who are
but if a CEO wants things to be a certain way, even if it's THE WORST idea, u have to do it or u get fired.
I'll never understand why anyone thought it would be a good idea to have the people with the least relevant opinions/ideas be at the top and all the experts at the bottom of the power heirarchy...
If the machine cultists want to invent Robot Satan in their techno-Protestantism to scare them away from doing things, I’m fine with that so long as they stop doing things.
Making things that people want and that work is a fine way create a viable business.
VCs don't want a viable business. They want a multi-billion-dollar world-changing unicorn. If you have a viable business, they'll push it to become one. And then it won't be viable anymore.
Tech produces plenty of things that people want and even more or less works.
What they want, though, is to find something they can sell to 7 billion people, the way the internet went from barely anyone in the early 90s to damn near everyone. They think they can do that with any product.
I think the primary difference and why nVidia is the current most valuable company on the planet speaks to a little more certitude this time -
The value is clearly there for companies, or simply put, they’d stop spending. Not propel them to the top of the F500.
That’s all to say LLMs have sucked all the air out of the room but ML in general has convinced everyone with “the most” of some kind of data to make some models.
I also expect they’ll “keep working” but reach a financial crux: If OpenAI spends 5billion on a model and it sucks, they’re done.
I think this is something that puts a lot more faith in their ability to behave as rational actors and not followers of a hype train than I’m willing to give them. admittedly, AI has more uses than say crypto did though, no argument there
Nvidia is selling shovels for back to back gold rushes. Their good times will end sooner or later but it will be a much softer landing for them than all the AI startups
Even in the window between the crypto meltdown and the start of AI madness, it was extremely difficult to get access to datacenter GPUs for boring ML at any moderate scale
although tbh, the emergent behavior of the market may be irrational, but the vast majority of actors are behaving rationally up to the limit of their computational power (very polynomially bounded) and their understanding of relative likelihoods (subjective to begin with and likely poorly updated)
Yeah. And my bit is “if you’ve taken your animal to the vet in the past two years you’ve benefited from AI/ML”. It’s just not rammed down your throat via marketing and FUD.
We often get lumped in with the Altmans of the world.
She likely wouldn’t see it yet if she’s outside the us and sometimes not even then as its somewhat “transparent”.
She likely has at least a few instruments from us that send the test results back to us though…
We then make medicine/clinical decision support recommendations thereon.
I'm not sure what you are referring to here. There are AI assisted tools for interpreting x rays but they're notoriously awful. I don't use and I don't know of any other vets that use ML/AI.
Also worth keeping in mind, I think, that “rational actor” for an individual isn’t necessarily rational for the org, and following a hype train is a great way to minimize reputational risk in the face of uncertainty. Chasing AI is the new not getting fired for choosing IBM.
I think Nvidia is built on a house of cards of timing. They were right place/right time, and now they're going to get hit with AMD/Intel starting to actually be competitive at the same time the hype cycle hits the trough of disillusionment. I think it'll be a bloodbath for them.
Years ago at a tech conference hosted by RBC one of Andreesens GPs was on a panel discussion w other VC and OPENLY said the dirty little secret of the VC industry is "our model is based on the greater fool" You just don't want to be last guy holding the hot Potatoe when the music stops
thinking of the ultrasound-based phone charger. this was impossible per literally physics, but VCs were quite explicit that they were getting in cos they were sure of greater fools.
In fairness, this did quietly happen in translation & localisation where a lot of professionals got turned into what a writers are worried about, being editors for the shit the machine translators pump out
