Clojurians Log v2

My position will most likely evolve but one thing is quite certain, I'm not going to spend a lot of money to feed the AI behemoths, I will be concetrating on local LLM's and processes where I can be quite certain of ROI.

I have much the same list as you, and I think the main problem is the forced "scaling" of LLM's. Local models are also built on the foundations laid by these companies, which still leaves a bad aftertaste, but at least inference is taken out of the data centers. Hopefully investor money soon runs out, before even more of the environment is destroyed by the continued push towards bigger and bigger models that we do not need (a naive thought, I know).

Yup, I feel somewhat annoyed because I feel that I do benefit from LLM usage and thus my benefits build upon achievements of those behemoths.

Not sure I understand the problem of AI from a fiscal point. Do you mean government spending on AI?

The only company currently making actual profit from AI is Nvidia. All others in the AI space are losing money/sucking up investment. (https://isaiprofitable.com) I wouldn't feel bad for OpenAI, et al. except for the fact that the investments these companies are making are creating environmental harm; they're benefiting from tax incentives provided by state and local governments in the face of vocal opposition from their constituents while benefiting relatively few (after the initial infrastructure outlay, there are relatively few jobs created by data centers); and the financial markets are coupled to these companies (think index and other funds) in ways that the potential losses incurred by these investments will be felt throughout the economy, not just for these companies individually.

@jeroenvandijk for example, Github cranked prices up a lot in August. This affects at least small companies quite heavily where all investments must have a solid ROI. And then the overall losses accumulating biz model of current Ai hype

Fiscally it's increasing the cost of electricity and ram and others which increases prices of consumer goods. Game consoles, phones, etc. So it puts a cost on everyone. And there's a lot of investment and pensions and insurance tied up into it, that if the bubble pops everyone will feel the impact and there could be tax payers bailout.

"Too big to f_ai_l"

True, good points. Then maybe I would like to add that I see non-IT friends around me in some kind of gold fever mode. Where they believe they can get rich quick after launching their AI prototype (someone more technical might just need to review it a little bit just to be sure..). Maybe the irrational hype captures this already

Outside programming and systems architecture the use case is most likely quite different and I cannot say whether it makes any fiscal sense or not. All Ai caveats apply there also.

I'd also like to add to repellants the intellectual theft that's it's built upon 😁

But ya, even if I agree to all the negatives and don't really like it, I also use it heavily and do find there is value in it, and if it's here to stay, it's likely disadvantageous for me to not at least take advantage of its upside.

I'm afraid our career paths do require AI skills from now on...

I'm having a hard time with this sense of AI inevitability. This stuff isn't like physics. I want to believe we (collectively) have choices here. Is this really a foregone conclusion?

We do, but is it a minority willing to go against the elite? It doesn't seem like there's really a global unity against AI. And it's likely not enough for even your own country to say regulate or outlaw AI use, if other countries do and can dramatically undersell, your companies might just be at risk, etc. To me it seems more similar to climate change, and we've barely managed to do anything to slow it down globally, because the competitive forces are too strong and getting universal consensus is too hard. And AI has an even smaller doomsday story, so I guess I'm just not hopeful haha. Sorry 😐

The problem with the inevitability argument is that it is presented as a fact, but it is an opinion. So many things were deemed to be ‘this way because they are’ and ended up being changed because people cared. So every time someone argues, “x is inevitable”, I mentally translate that to “x is useful to me, and the downsides to x are not severe enough to me to want to change them and I believe others to believe the same”.

That is not wrong, I just feel (emphasis on the word feel) that the majority in our industry has taken AI into their heart and it is hard to give up because all the promises were so sweet. I hope that in the end AI is treated as a tool and its ROI is delibaretely calculated befaure its usage is required. Also, I hope that interviews for positions where AI is in use actually focus on "how do you leverage AI in your work" instead of simplistic "can you use AI?", that is, nuances, pitfalls and caveats inherently within AI are taken into consideration.

Same; a more nuanced debate on the value + downsides, is beneficial in my opinion.

But for me, that means trying to steer away from … I don’t know how to word this… such hash arguments as ‘this is inevitable’ because those don’t allow this nuance.

If you're a podcast listener, Ray and I shared some thoughts on the inevitability narrative here: https://politechs.dev/episodes/ai-05-inevitability/

What is certainly inevitable is the environmental destruction that will be wrought by staying on the current path.

This is a post on how there is a cognitive bias when thinking about code as assembly language and developers don't need to look at it. Rather than challenging the assertion on technical arguments, I try to highlight the cognitive bias and compare it to other fields (mathematics etc).

Well said @U0BFT9VHWVB.