I don’t apply generative AI or LLMs in my work at this time. You can trust that artifacts that I author (prose or code) haven’t involved them at any stage. I haven’t built up a coherent dogma about these tools yet; I’m trying to follow the field very closely and I’m overall very ambivalent on the topic.
I do use tools that include Whisper audio models.
Per a variety of metrics, the conventional resources I’m using now seem to be working reasonably well. That said, I do believe that as a taxpayer-funded researcher I have an obligation to the public to try use the best tools I can in my work, not just those I happen to already know or be comfortable with. Trying to find the global maxima rather than a certain local maximum, “one’s reach should exceed one’s grasp”, etc. So I admit that this is bit of a cop-out.
In the past, I’ve seen people kibosh new tools in ways that I think are not reasonable - like the furor around 2010 about how Wikipedia was not an acceptable source, or examples like telling off a grad student for watching a training video rather than reading a textbook on a specific topic. I don’t want to perpetuate these kinds of traditions.
However, this is my internal position at this time.
I feel like part of how I learn comes from little unscheduled diversions in the course of a task - from muddling inexpertly through small projects into larger things - and I feel that, compared to other tools, AI could have a larger risk of corrupting the ineffable parts of that process in a way I won’t recognize.
Conversely, I’ve encountered reports of people using AI to teach themselves things. This January I met a woman at conference who moved from basic analog electronics into the field of semiconductor detector readout design. She said she taught herself detector physics partly using textbooks, but also using Generative AI to ask questions about book content and getting it to make tests and flashcards etc.
An overall summary might look like this:
[!note]
There’s a lot of discourse around GenAI in the circles I run in. A lot of it could be easily dismissed by a casual engagement - it initially sounds absurd to ban someone from contributing to a software project using a tool for environmental reasons. But, as one of the smartest and extremely even-keeled people I know put it: any activity that consumes fossil fuels needs to be considered with trepedation. In that context, it really does make sense.
yet I feel somehow that a lot of these proximate effects of genai are… kind of more a knock-on effect of the, erm, overall collapse of western civilization we’re experiencing?
- I would argue that the reason why the energy use is a problem is because carbon dioxide and greenhouse gases were just deregulated as pollutants by the EPA. The fact that a new, power-hungry industry has sprung up is kind of a footnote in that story; if e.g. microsoft hadn’t been allowed to scale back its committments.
- The reason why it’s a problem that textbooks are being bought, scanned, shredded, and then the scans deleted, is plausibly because of the atrophy of public libraries and the failure of governments to build systems that meet the public’s demand for trustworthy information. Okay, buying all the books and shredding them is a douche move anyway, but it wouldn’t be so catastrophic if there was an equal effort to preserve them.
- The scraperbots are, bizarrely, to some degree a consequence of Ukrainian ISPs being forced to sell IPv4 blocks to residential proxy companies.
[!danger] -
The scraperbots. Generative AI companies are buying residential proxies and are direct, intentional, negative impacts on any site that isn’t behind a CDN. lwn.net, and
[!note]
that has had downstream impacts. almost unmanageable burden on sysadmins.
[!note]
this effort to bar automated access comes at pretty much the worst possible time, since the good guys need fast programmatic access to archive and rescue data from the collapse of the US funding infrastructure, and so there’s this three-way arms race going on.
[!note]
A colleague of mine just told me that he basically is able to code one day a month, because he uses up tokens.I think it would be impossible for me to work this way and cements my idea that, with the current limitation in computing power, at this moment the commercial AIs are one-armed bandits.
[!success] +
2. Don Knuth, Terence Tao, and his colleagues find the outputs novel and interesting, which is very persuasive to me.
[!danger] -
When people use an LLM as an intermediary in a conversation, it seems to often breach social contracts and norms in a way that makes it hard to engage with.
You can frequently see this in Github issues, for instance. A notable example would be some email technical support I got from a company.
[!note]
The scale of capital expenditures on AI infrastructure is completely unfathomable to me… and there’s also an element to this that feels kind of disrespectful.Climate scientists use supercomputers to evaluate earth-system models, the outcomes of which reflect how climate change will affect everyone on the planet.
The return on investment of climate mitigation efforts is known to be between 2-3x, a return which is all but guaranteed and which does not rely on market interest, and the results of models are prerequisite to mitigation.
One computer dedicated to this task is the NWSC-3 (Derecho) supercomputer, which was specced out in 2018, RFP’d in 2020, came online at the end of 2023, and cost $40 million overall.
The people who were involved in getting Derecho are not less competent or motivated or efficient workers than those at xAI.
Imagine you gave $500 billion to systems librarians, with the mandate to build systems and resources to answer people’s questions, and provide a legal basis to bypass copyright.
