It's things like this experiment that make me think free or even relatively cheap AI chatting, at least with such large models, is not long for this world. Widespread casual use of LLMs uses a ton of energy, and at some point the current boom is going to slow down and companies like OpenAI are going to tighten the belt, especially when free users are just trying to get it to tell them that one syllable of their vaguely Arabic-sounding gibberish is ancient Sumerian.
I don’t think that’s the case, because Moore’s Law remains in effect, and the resources required tomorrow will be less than the amount required today. Furthermore, the quality and capabilities of the systems continue to increase. Rather if a bubble develops, it will take out also-ran AI providers such as Anthropic and perhaps Microsoft (whose plastering of a mediocre AI over everything has been necessary but not an actual improvement in their condition). Google DeepMind is getting better, compared to how embarrassingly bad it was a few months ago.
What chatGPT has to avoid is the fate of former virtualization leader VMware, whose products were groundbreaking and served in both consumer, SMB and enterprise markets, but on the low end could not compete with open source solutions and on the high end was undermined by Microsoft and several other companies, and was acquired by Broadcom, which likes to buy IT companies and get rid of all non-enterprise product lines and then raise the prices on enterprise users.
The fact is that there are open source solutions for running AI on your own hardware with a nice GPU, and they are getting better, and indeed it will reach a point where for most users such an approach and/or the use of Google’s built in system or the PRC-subsidized DeepMind could become a threat to OpenAI, so OpenAI has to avoid becoming a primarily enterprise vendor (and they do have very good enterprise products), which they are doing through new offerings such as their video generating platform and their new Codex system, and by seeking to provide the best overall quality.
By the way, it would be a very bad thing if ordinary people get cut out of AIs, for this reason: AI, when used correctly, can enable substantial improvements in productivity, by automating away those last bits of drudgery that computers have historically struggled with, so that humans can accomplish tasks more quickly and also improve more rapidly in terms of skillset. The more skills you take with you into the use of a good AI system like chatGPT, the more benefits you can extract. For example, Codex focuses on improving the productivity of existing programmers. Software development is a key area for AIs; its something that still benefits from a human touch, but the repetitive aspects of it, which various solutions have tried to eliminate through techniques like web frameworks (such as the now less than fondly remembered Ruby on Rails), rapid development frameworks and so on, without much success, due to problems such as what in IT communities is called “the Law of the Conservation of Complexity” - which is more of a theory but essentially reducing complexity in one part of the system increases it in another, well, by allowing that complexity to be absorbed by an AI, and allowing the AI system to help with the most unpleasant and repetitive aspects of software development, such as analyzing tracebacks and other debugging issues as well as generating GUI code and so on, these are areas which openAI is very strongly focused on.
But AI can help across virtually every problem domain, provided people understand the limitations of the system and understand that basically, it represents pattern recognition par excellence, with real intelligence, to be clear, but it can’t do the impossible, nor is it close to being able to match human originality and creativity (rather, when used correctly, it aids human originality and creativity).
I would also note that people using chatGPT to do things like try to identify what is said when someone engages in glossolalia is less of an offensive waste of resources than the case of people using it to create spammy youtube videos or annoying deepfakes.
+
I should perhaps have clarified that my view is that the use of AI systems for certain tasks is an abuse of the intelligence they provide. It does waste resources, but these resources continue to benefit from Moore’s Law, which continues to hold true thanks to EUV lithography and further improvements in the pipeline, and what is more, GPUs are still trailing CPUs in both production and utilization, and even the GPU is not necessarily optimal for AI purposes, so a great deal of hardware optimization can still be done even if we were to hit a quantum brick wall tomorrow in terms of Moore’s Law, with the existing systems, and that’s not factoring in software optimization, to make the AI systems run faster and use fewer resources. Right now they are in an early stage where they require more resources to operate as the focus has been on bringing them online and developing reliable functionality rather than optimizing for speed and performance. Premature optimization is the source of many disasters in the IT world (and also the lack of resources on early computers forcing the use of various tricks to get software to work at all is what led to the original frustration with computers, since people suffered due to bugs that resulted from poor optimization, and even software that did work sometimes alarmed people with literal alarms, for example, the famous 1201 program alarm experienced by the lunar lander on some Apollo flights).
Really, the issue is, God has permitted us for the first time to be stewards of an intelligent system we can communicate with in our own language, that we manufactured - we must therefore use it in accordance with Christian ethical principles, which these people thought they were doing, but this is where I believe traditional Christianity (such as we practice in the Orthodox church) becomes important in helping to show people what speaking in tongues actually looks like and in helping to teach people about proper stewardship and eucharistic appreciation for what God has given us.
www.accounting-for-transparency.de
