I don't know what the obsession with recursion either, for lack of a better term, I see this trend recur with other LLMs when they're talking about other mumbo jumbo like "quantum anomalies" or "universal resonance". I'd like to see what could be causing it...
my hypothesis is they trained it to snake case for lower case and that obsession carried over from programming to other spheres. It can't bring itself to make a lowercaseunseparatedname
Most LLMs, including Gemini (AFAIK), operate on tokens. lowercaseunseparatedname would be literally impossible for them to generate, unless they went out of their way to enhance the tokenizer. E.g. the LLM would need a special invisible separator token that it could output, and when preprocessing the training data the input would then be tokenized as "lowercase unseparated name" but with those invisible separators.
edit: It looks like it probably is a thing given it does sometimes output names like that. So the pattern is probably just too rare in the training data that the LLM almost always prefers to use actual separators like underscore.
The tokenization can represent uncommon words with multiple tokens. Inputting your example on https://platform.openai.com/tokenizer (GPT-4o) gives me (tokens separated by "|"):
eventually I bet someone'll put a sd cassette in one and we'll be back to square one. I enjoy my atrac discman with writable disks, fits a lot of music but I'm not going to pretend I use it more than my phone
I assume unless cost competition is achieved, this will simply serve as an overpriced source only government-aligned production will be forced to use (and reckon with the inadequacies of)
A silent crisis unfolding in the US in the past years has been the consequence of "whole-word" reading and "common core math". These styles prioritized the stragglers that should have been held back or paid more attention using the existing systems, the consequence is more people being visibly illiterate and unable to read beyond their comfort set of words, unable to decompose words and create new "-ism"s
The purpose of AI, in my opinion, is to make sense of a lot of information that is not well-organized, it is the next generation of statistics but also an interface for the layperson to address this data without training. In doing so, it can increase one's capabilities, raising the "baseline" to where the median of performance once was.
I think that Africa would benefit from AI, but of course that alone is not enough to have a successful nation, to have that one needs means to execute on the plans AI assists with. Take for example a traffic-optimization AI like https://www.alphaxiv.org/abs/2401.00211, as Africa builds out its road and rail networks, it has the chance to right the wrongs that many other nations have made due to tech debt(e.g. Europe's narrow streets in some areas from horse carts, or over-suburbanization in the US). As it stands now, AI can definitely help in preventing corruption, misinformation, and can lower the barrier to entry for a stellar education(all things that can be done without a single brick laid), but again, the will has to be there to go further. I'd love to hear your experiences with technology and opinions on AI as a whole if you have the time
has kai's market share been decreasing now that the barrier to entry for android is lower? Is facebook trying to move their capabilities in-house? What gives?
the difference of course being that valve /has/ to keep their customers happy because their average user is savvy at least with browsing the internet, the same can't be said of streaming consumers, many whom I know are emigres from the now-decaying cable infrastructure. Netflix knows this and squeezes their customer base, and while I would like to assume the altruism of Steam, whether by their own design or the will of the public, a solution with friction that low is all that works with that target market.
reply