🚀 I used to kill bugs at Microsoft, now i run a startup focused on making $BRK irrelevant w/ the world's best #ML 🤖
#digitalhealth 😍 #dataisKING HealthOS
Sam Altman compared the AI compute shortage to people hoarding toilet paper during COVID, then said something worse. If models keep getting smarter and cheaper, the shortage never ends.
This is the part of the AI story most people get backwards.
Normally, when something gets cheaper, the shortage goes away. Prices fall, supply catches up, the panic ends. That is how almost every market works.
Altman says AI breaks that rule.
He confirmed there is a gigantic compute shortage right now. H100s are basically gone for the year. The interviewer compared it to COVID, when all the toilet paper vanished off the shelves. Altman didn't argue.
Then he went further.
He said if you make models smart enough and cheap enough, demand is essentially uncapped. Every price drop unlocks more uses. More agents running for you. More tasks you hand off. The cheaper it gets, the more you want.
So the demand grows faster than the supply ever can.
His point was simple. As long as AI keeps improving, there will be a shortage forever.
Most people think falling prices end shortages.
Altman is saying this is the one case where falling prices cause them.
(Watch the full talk on YouTube at Stanford Online channel)
Nadie en Google fue imbécil lo mismo que nadie en Kodak fue imbécil.
Cuando tienes algo que defender tu cerebro entra en modo protección, lo que normalmente desemboca en tu muerte, pero en su momento evolutivo tuvo sentido.
Que una empresa haga limpieza cada ciertos años es de lo más sano del mundo, especialmente si te quieres mantener competitivo.
Gente improductiva se acumula y los engranajes empiezan a ir cada vez peor… el rollo de las familias está muy bien, pero una empresa no es una ONG.
Pre-orders for Grand Theft Auto VI will officially begin on June 25 on digital storefronts and at other select retailers.
Check out the official cover art, also available as downloadable artwork at https://t.co/XPwC8URCQ4
A few thoughts that come up from this suspension of Fable/Mythos:
1. Even if this decision is changed the move set a precedence. Investors investing in AI model providers will start to price for this risk. The idea of one company monopolizing and developing “digital god” and reaping all the economic benefits of it is dead. This opens the question of the ability of these top frontier AI labs to pay for the big compute commitments.
2. Enterpises will rush to orchestration platforms as they will want to have the option to switch between different AI model providers with ease not just because of effectivness of models but also because of the risk that the model or provider is banned by the government. Hyperscalers are in a great position to be that orchestration layer between different providers.
3. There is going to be an acceleration of AI models exclusive for specific vertical use-cases (models that dont have cyber or other capabilities so the chances of gov reach is lower), competiton here is going to expand to outside of the top 3 AI labs IMO.
This is a big turning point for AI regulation.
The government is starting to deem some models too powerful for certain uses, which creates a precedent for a range of possible controls in the future.
I’m in the camp that this is unnecessary and we should be primarily regulating the use of AI, as opposed to the underlying models. But, equally, there are plenty of people that actually prefer this outcome.
Either way, it’s unlikely that we’re going back to a world where the government doesn’t have far more meaningful involvement in the rate of AI progress.
There’s no amount of intelligence that can get packed into AI models that replaces the need for context. For any sufficiently general purpose AI, you will always have to guide it in the direction you want as it has an infinite range of directions it can go in.
As long as the same model is used by a lawyer, an engineer, a financial analyst, or a healthcare professional, and as long as you’re trying to do anything uniquely differentiated or specific, then instructions, domain context, and proprietary data will always need to get into the context window for the model to be useful.
This is partly why AI automation doesn’t come for free, and why there’s still a wide spectrum of who’s getting the largest gains from AI and who’s not. You have to put in real work, and you get real value on the other end.
This is one of the advantages that applied AI will also have in the market. Any layer of abstraction above just the raw intelligence that can meaningfully get you off to the races faster will likely continue to be valuable.
Token costs are becoming one of the hottest topics for any enterprise I talk with right now. It’s very bullish for AI in general because it means these systems are being used at a scale that wasn’t contemplated before.
It also gives way to another form of differentiation that will emerge for the applied AI layer, which is model routing.
As tokens take on a significant amount of the cost of any given workflow, then companies will inevitably want to ensure that their dollars go into the most efficient use of tokens for the particular job at hand.
Frontier intelligence will always be relevant at the high end of tasks, like coding, legal and financial analysis, healthcare, and more. And dollars spent here will only go up over time. But, equally, you can peel off individual tasks to lower cost models (whether they’re from open weights vendors or the major labs) and deliver a more efficient end outcome.
To do this effectively, the applied AI layer needs to understand the workflows in their domain better than anyone else, and be able to mix and match models to different jobs. If you’re doing document extraction, you need to know which models perform better or worse for any given document type. If you’re legal analysis, you want to know which models perform various types of tasks best. And so on.
This will become one of the bigger differentiation points over time. The companies with the best evals, the best ability to route the workloads, and those that have business models directly aligned to customers financial goals, will be in a great position.
I want some kind of LLM workflow tool.
• Ability to manage a set of input files (Markdown or similar), plus other general-purpose context.
• With real-time collaboration. (And maybe some concept of snapshots or VCS integration.)
• And the ability to create/manage a inference workflows and a stored set of prompts.
• Access to general-purpose coding agents (and not just chat models).
• Some concept of compiled outputs/inference results (which ideally can be shared externally).
Many projects have this feeling: "there is all this stuff, which I want to process/compute over in this iterated way, with some build artifacts being important/worth saving." GNU Autotools x Notion or something. Is anyone building this?
@pepe_sepulveda Mollick lo resume perfectamente:
IA como muleta → destruye aprendizaje.
IA como tutor socrático (prompted para que te haga explicar, razonar y adaptar) → efectos positivos claros en ensayos aleatorizados.