A.I. Joe (@A_I_Joe) header
A.I. Joe (@A_I_Joe) profile picture

If you can't beat them, build them. #OpinionsAreMyOwn

Auckland, New Zealand
Joined November 2013
Prompting #LLM's with 1M words with @MikhailBurtsev youtu.be/4QDvNiyAikw We interview Dr Mikhail Burtsev, author of the paper, "Scaling Transformer to 1M tokens and beyond with RMT". Also available on your podcast platform #contextwindow #nlp #AI #ML #ChatGPT
1
1
2
666
Visualization of Git CLI Command Effects complete-broom-c1e.notion.si…
1
104
Re-Evaluating GPT-4's Bar Exam Performance by Eric Martínez :: SSRN papers.ssrn.com/sol3/papers.…
105
Right now #AI (like recommendation engines) are competing for your attention, to keep you watching. Next gen's algorithms will compete for your heart, seeking to become emotionally intimate with you, so it can persuade you to do, well, anything.youtu.be/xoVJKj8lcNQ #LLMs
1
1
88
This isn't because it will "want" to, it is simply because competitive economic forces will drive large corporations to optimize it to do this. It's a pretty compelling argument to regulate.
66
#Teachers are making AI detectors literal automated gatekeepers of human academic and career success, with no opportunity for recourse, and that's got to be illegal according to something #chatgpt #AIDetection
86
The recent ChatGPT panel, featuring Nickle Lyu, Dr Alyona Medelyan and me, is now on Youtube: youtube.com/watch?v=Lfda3ijv… Also available in audio format on: - Spotify: open.spotify.com/episode/6aw… - Apple Podcasts: podcasts.apple.com/nz/podcas… - Your favorite podcast platform #AI #ChatGPT
1
65
ChatGPT Gets Its “Wolfram Superpowers”!—Stephen Wolfram Writings writings.stephenwolfram.com/…
57
Within 5 years, the major health concern of people spending too much time in customized alternative reality simulations will be a trending headline #future
46
You can safely ignore all the sensational headlines you'll see about #ChatGPT being #sentient. It doesn't understand "meaning" in the way we think of it. The reality is that it literally never knows the next word it's going to say.
1
1
105
For every word it says, it just looks at everything that came before and determines in a syntactically and symbolically-aware way what would be the most likely word to come next given the average content it's seen, and spits it out.
1
61
It makes no value judgements about it, and all it "thinks" about is how likely the next word would be on average, and says the most likely. Anything beyond that has to be imposed on the model by external manipulation
55
Heads-up for anyone who heard about #ChatGPT, but dismissed it as just a another new chatbot. ChatGPT is a new type of chatbot in about the same way the first Ford motor factory was a new type of wrench. It's going to transform most industries
1
1
127
Dark mode for Notepad? Why didn't you lead with that Microsoft? I would have installed #Windows11 ages ago.
1
7
I don't much like the idea of teams owning microservices (or services used as a way to decouple teams from each other). I'd prefer to use bounded contexts for that. Teams work on stories—full end to end implementations of a narrow vertical slice. 1/4
17
40
5
332