Ryder · @Ryder
95 followers · 489 posts · Server sigmoid.social

based Machine Learning won’t threaten humanity- that’s clear by now.

But irresponsible usage by greedy corporations like will produce large amounts of toxic waste that will make our lives more miserable.

#transfomer #openai #microsoft #anthropicai #google #meta

Last updated 1 year ago

Giskard · @Giskard
65 followers · 106 posts · Server fosstodon.org

It was a pleasure to be part of this event and witness the come together, especially on topics as critical as and LLMs' security. 🛡 🙌 We had the opportunity to engage with leading AI teams from
@huggingface, @stabilityai, , and
[2/5]

#aicommunity #aisafety #cohere #anthropicai

Last updated 1 year ago

Victor Paléologue · @palaio
50 followers · 143 posts · Server fediscience.org

Finally, I have tested !

I felt like it was hallucinating less, was more eager to acknowledge when it did not know an answer and was more accurate.

explained general relativity’s twin paradox in a consistent manner whereas would contradict itself.

can generate code to drive a robot, including animations!

So far, looks like a good alternative, and to me have a more reassuring ethical stance than .

#openai #anthropicai #nao #python #chatgpt #llm #claude

Last updated 2 years ago

· @Techronic9876
115 followers · 1895 posts · Server sigmoid.social

One interesting thing about ’s LLM is that it has a different “voice” than which seems somehow less annoying

ChatGPT has a lot of mansplaining vibes that doesn’t have. Claude also makes a few typos which makes you feel you can trust the non typo text more.

#anthropicai #chatgpt #claude

Last updated 2 years ago

· @Techronic9876
114 followers · 1890 posts · Server sigmoid.social

Playing around with Claude and it seems to be very fast and high quality. Definitely as good as gpt 3.5 in a few quick tests. OpenAI does a good job formatting the markdown but I’m sure anthropic UI will catch up.

#anthropicai

Last updated 2 years ago

Ryder 🕯️#UnplugAI · @Ryder
74 followers · 647 posts · Server sigmoid.social

reports that very large can be steered either way (biased/less biased) by prompting, corroborating the findings around evil - prompting is all you need.

arxiv.org/abs/2302.07459

#anthropicai #llm #bing

Last updated 2 years ago

Alhaythum · @Alhaythum
33 followers · 65 posts · Server noc.social

@googlecloud @financialtimes @cointelegraph

also confirmed that they have raised capital from & his former partners from & The Center for Emerging Risk Research in Series B round.
shows that the fundraising by announced on April 29, 2022

crunchbase.com/funding_round/a

#crunchbase #cerr #AlamedaResearch #SBF #anthropicai

Last updated 2 years ago

Alhaythum · @Alhaythum
33 followers · 64 posts · Server noc.social

@googlecloud invests $300M in AI firm previously funded by
The $300M figure reported by @financialtimes, while confirm the investment partnership with @googlecloud without disclosing any figures

Via @cointelegraph

cointelegraph.com/news/google-

#ftx #SBF #anthropicai

Last updated 2 years ago

Arie van Deursen · @avandeursen
320 followers · 119 posts · Server fediscience.org

“Why is it that neurons sometimes align with features and sometimes don't? Why do some models and tasks have many of these clean neurons, while they're vanishingly rare in others?

In this paper, we use toy models — small ReLU networks trained on synthetic data with sparse input features — to investigate how and when models represent more features than they have dimensions.“

transformer-circuits.pub/2022/

#superposition #interpretableai #anthropicai

Last updated 2 years ago