What constitutes real "thinking" or "reasoning" is beside the point. What matters is what results we getting.
And the challenge is rethinking how we do work, connecting all the data sources for agents to run and perform work over the various sources that we perform work. That will take ages. Not to mention having the controls in place to make that the "thinking" was correct in the end.
Thinking is not besides the point, it is the entire point.
You seem to be defining "thinking" as an interchangeable black box, and as long as something fits that slot and "gets results", it's fine.
But it's the code-writing that's the interchangeable black box, not the thinking. The actual work of software development is not writing code, it's solving problems.
With a problem-space-navigation model, I'd agree that there are different strategies that can find a path from A to B, and what we call cognition is one way (more like a collection of techniques) to find a path. I mean, you can in principle brute-force this until you get the desired result.
But that's not the only thing that thinking does. Thinking responds to changing constraints, unexpected effects, new information, and shifting requirements. Thinking observes its own outputs and its own actions. Thinking uses underlying models to reason from first principles. These strategies are domain-independent, too.
And that's not even addressing all the other work involved in reality: deciding what the product should do when the design is underspecified. Asking the client/manager/etc what they want it to do in cases X, Y and Z. Offering suggestions and proposals and explaining tradeoffs.
Now I imagine there could be some other processes we haven't conceived of that can do these things but do them differently than human brains do. But if there were we'd probably just still call it 'thinking.'
> connecting all the data sources for agents to run
Copilot can't jump to definition in Visual Studio.
Anthropic got a lot of mileage out of teaching Claude to grep, but LLM agents are a complete dead-end for my code-base until they can use the semantic search tools that actually work on our code-base and hook into the docs for our expensive proprietary dependencies.
You have to understand that these large corpos move like whales, and the money you quoted is a rounding error. I´ve seen a company department burn cash it was asigned on purpose so it wouldn´t go back to finance (indicating that the department isnt using all their money and something is wrong).
Making the mental adjustment coming from a public college to a private company regarding money has been one of the most difficult transitions I've ever faced.
At the college I had a $250,000 annual budget for IT that had to be planned out to the best of my abilities nearly a year in advance of the actual physical year, get approval from 2-5 levels of management for the budget, and then be flexible on when the money became available to purchase depending on the states fiscal economic factors.
In a private company now, even when I volunteer to do something to save money, they say it's not worth my time and effort, pay someone to do it.
Purchases under $25,000 can be made without approval, over that I just have to ask my direct report for approval.
I'm still personal finance budget minded, so they don't have to worry about me buying gold plated toilet hinges or anything, but it's still financial whiplash even years later for how they do things.
Nobody with interest in politics thinks it's about drugs. It's a pretext and a way to gain legitimacy to exert force over foreign nation with some legitimacy that would otherwise clearly go against international law.
Most of the USA's refineries specialize in low grade oil. The best grade oil is often shipped out of the USA for refining. Shipping costs are so low on a grand scale that it's more profitable to ship the USA's high quality oil overseas than building new refineries in the USA just for that:
https://www.marketplace.org/story/2024/05/13/the-u-s-exports...
Most attention implementations can work across an arbitrarily long context.
The limiting factors are typically:
1. Often there are latency/throughput requirements for model serving which become challenging to fulfill at a certain context length.
2. The model has to be _trained_ to use the desired context length, and training becomes prohibitively expensive at larger contexts.
(2) is even a big enough problem that some popular open source models that claim to support large context lengths in fact are trained on smaller ones and use "context length extension" hacks like YaRN to trick the model into working on longer contexts at inference time.
The model will use the full context if it's been designed well, but you can still increase the size of the window on models where it hasn't. It's just pointless. People who don't know much about LLMs will still think "bigger number is better" though.
There is some circular financing going on, but AI accelerationists think this will be offset by demand, value, and adoption in businesses. Hence these deals are warranted for the incoming demand.
I think I saw Altman saying there's a global shortage of compute just now so this may address it. I'm not sure how much is actual user demand though and how much just investors wanting to pile into AI startups.
>> "Turns out the major bottleneck is not intelligence, but rather providing the correct context."
But this has more or less always been the case for LLMs. The challenge becomes context capure. Which in my opinion is the real challenge with LLM adoption.
Without the right contex, some tasks just cannot be reliably completed.
This concept is closely reated to politics of inevitability coined by Timothy Snyder.
"...the politics of inevitability – a sense that the future is just more of the present, that the laws of progress are known, that there are no alternatives, and therefore nothing really to be done."[0]
This is incredibly useful and interesting. I have tried asking claude to generate XML code for various diagrams for import to draw.io with varying success. But I feel like if I could incorporate these instead, or markdown, for a specific graph instead of pure XML would yield better results.
I think the better question is to answer why do emergent properties exist in the first place.
I disagree with the premise that emergence is binary. It's not. What we determine "emergent behaviour" is partly a social concept. We decide when an LLM is good enough for us and when it "solved" something through emergent properties.
And the challenge is rethinking how we do work, connecting all the data sources for agents to run and perform work over the various sources that we perform work. That will take ages. Not to mention having the controls in place to make that the "thinking" was correct in the end.
reply