~/talk

Mirror, mirror: LLMs and the illusion of humanity

on YouTube 58m
Talkat PyCon Italia 2024 · 2024-05-22..05-25· [large-language-models, machine-learning]

speaker

name:Jodie Burchell
github:t-redactyl

resources

video:https://youtu.be/4jGwEOGunrU
slides:none on file

summary

Covers specific high‑profile claims and tests: GPT‑3 and GPT‑4 produce Shakespeare‑like text and exam‑passing outputs, Microsoft researchers claimed “sparks of AGI”, Google’s Lambda prompted Blake Le Moine to claim sentience but company review found evidence against sentience. Demonstrations include word‑embedding analogies, the Octopus thought experiment, and a Codeforces test where GPT‑4 solves 10 training‑era problems but fails 10 post‑training problems, showing memorization, contradictions and brittle generalization rather than integrated, grounded understanding.

AI-written summary · about

chapters (6)

  1. 0:00Hype and misconceptions
  2. 3:44LLMs vs human qualities
  3. 10:39Language and semantics
  4. 26:52The myth of sentience
  5. 40:05Intelligence and generalization
  6. 53:38Conclusion and ethics

related talks

~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.