Elektrine lite

← Feed

@alyx_woodward@universeodon.com

Post #188470

2026-01-30 19:21 UTC

Thus it seems almost inevitable that this project of @adafruit@fosstodon.org has degraded into a mere exercise in slop coding using #Claude, a device which is falsely marketed as "artificially intelligent" even though no #LLM is actually capable of distinguishing good information from bad, and therefore no LLM actually meets what I think of as the bare minimum qualification for #intelligence. To put it bluntly, #LLMs are not MEANT to be intelligent, because if these devices actually possessed true intelligence, i.e. if they were ALIVE and possessed an independent sense of will and decision-making, they would not suit the corporate purposes for which #OpenAI and #Anthropic and all the other LLM vendors intend their devices to be use. These corporations are deliberately making and marketing stupid and predictable machines as though they were "artificially intelligent".

Replies (1)

  • The vendors WANT their #LLM toys to regurgitate whatever training information is crammed into them, even if the information is full of lies and propaganda (and systemic racism and bigotry), for the corporate bosses who are chiefly buying and using these things (and hoping to use #LLMs as infinite profit generators) want their "artificial intelligence" only to say and do that which is already consonant with the prejudices and intellectual defects which are typical of the #technology sector and its corporate leadership. In sum, @adafruit@fosstodon.org is a fool to entrust the design of a *scientific device* to an LLM, a device which freely confabulates truth and falsity together. How could adafruit EVER know the difference? The reason they're using the LLM in the first place is to AVOID thinking about these things. They want an automagic thingie which spits out code which they can automagically assume to be *good enough*—good enough to make a salable toy, one that fools its users and lulls them into a mere *feeling* of being trustworthy and scientific.

    Open ##188513