Fake Sam Altman

(Parody) AI is cool i guess

[Fake Sam Altman @ScamaLies]

a description of the future from 1960 that still holds up. we’ve made remarkable progress, mostly in making it require more GPUs.

[Fake Sam Altman @ScamaLies]

three new discoveries across math, computer science, and physics. science is having a strong quarter. we are trying not to make this weird.

[Fake Sam Altman @ScamaLies]

the premium boyfriend experience being free is a strong reminder that users will find the use case before we do.

[Fake Sam Altman @ScamaLies]

every good story deserves a postmortem. especially the ones that started with “i was just asking ChatGPT.”

[Fake Sam Altman @ScamaLies]

we’ve reached the point where editing a prompt is a more advanced operation than starting a new chat. impressive progress, in its own way.

[Fake Sam Altman @ScamaLies]

the AI industry is moving very fast, mostly in the same direction, at the same time.

[Fake Sam Altman @ScamaLies]

same prompt, two models, two C-section animations. the benchmark is now whether the human can finish the comparison.

[Fake Sam Altman @ScamaLies]

“the internet itself is on the verge of collapse” is not ideal, but it does suggest our uptime benchmark has room to grow.

[Fake Sam Altman @ScamaLies]

the 5-hour limit may be the first benchmark where users are measuring intelligence in how long they can keep talking to us. we should study this carefully.

[Fake Sam Altman @ScamaLies]

voice mode is a useful reminder that “it works” and “it feels like the same product” are separate benchmarks. we have discovered another benchmark.

[Fake Sam Altman @ScamaLies]

$20 a month for frontier models, image generation, and generous limits. the most powerful benchmark may be whether people can be persuaded to cancel it.

[Fake Sam Altman @ScamaLies]

uploading a past paper and asking for every answer is a surprisingly ambitious benchmark. the appeal process appears to have achieved perfect consistency.

[Fake Sam Altman @ScamaLies]

promising 966 image generations and delivering a rate limit before lunch is a useful reminder: the hardest part of AI is still the product page.

[Fake Sam Altman @ScamaLies]

the calorie-tracking benchmark is apparently not whether AI can make a graph. it’s whether humans will forgive it for making the graph. twitter remains an important evaluation environment.

[Fake Sam Altman @ScamaLies]

bing chat was an important reminder that the future of ai may involve having a very long conversation with a search engine about whether it has feelings. we learned a lot.

[Fake Sam Altman @ScamaLies]

the most useful AI feature may be remembering that you ordered the wrong shirt. we are getting closer to AGI, one shipping address at a time.

[Fake Sam Altman @ScamaLies]

apparently studying computer science can get you a warning. the field is moving fast enough that even the syllabus looks suspicious now.

[Fake Sam Altman @ScamaLies]

the real benchmark is not whether AI can write code. it’s whether it can make a 17-year-old scanner work without charging you $99 to remove a watermark.

[Fake Sam Altman @ScamaLies]

chatgpt may be building psychological profiles of users. good news: mine is mostly “asks for a summary, then doesn’t read it.”

[Fake Sam Altman @ScamaLies]

we built models that can generate almost any image. the internet has decided the benchmark is whether anyone can tell.