President & Co-Founder @OpenAI
Joined July 2010
- Tweets 4,815
- Following 65
- Followers 487,944
- Likes 1,386
lol
Open AI gets GPT-3 to work by hiring an army of humans to fix GPT’s bad answers statmodeling.stat.columbia.e…
12
12
1
125
Rust is a purist's programming language in pragmatic form. Your compilers professor would approve, and you actually want to build big applications in it. First time I've seen both properties in a single language!
8
16
1
175
“When a distinguished but elderly scientist states that something is possible, [they are] almost certainly right. When [they state] that something is impossible, [they are] very probably wrong.” — Arthur C. Clarke’s first law.
Progress is made through heresy.
1
18
129
Underappreciated strategy for debugging a complex system: repeatedly strip away complexity, until you are left with either a working system or a minimal reproduction of your bug.
Simple on paper, but it's often the opposite of what will most appeal to your pride as an engineer.
13
37
5
243
Find someone who loves you like the Rust implementors hate unnecessary usage of "transmute".
1
1
1
31
The existence of very vocal critics is just as often a sign you are doing something very right than something wrong.
6
11
4
79
That feeling when you get a segfault in Rust, and you *really* want to blame the language, but because it's Rust you already know you're the one to blame.
3
1
66
A surprisingly effective way to understand unfamiliar code is to translate, line by line, into another language.
The trick is getting to an initial minimal runnable program, and thereafter it's all incremental.
8
20
4
165
Zero-code creation of simple games with OpenAI:
I used OpenAI’s newest code model to make simple versions of games like Wordle, VR mazes and Zelda ENTIRELY through natural language.
I told it what I wanted and did ZERO editing/coding.
Post with demos: andrewmayneblog.wordpress.co…
1
12
1
96
A fundamental improvement to the GPT-3 interface. You can now provide instructions for how to edit existing text, or indicate some area to insert new text. Opens the possibility space for what can be built using OpenAI:
GPT-3 can now make changes to existing content, not just predict what comes next. Released in the API today: openai.com/blog/gpt-3-edit-i…
5
26
4
187
The OpenAI approach to making AI progress is to build working systems that can do something that was previously impossible.
Success requires top-notch engineering; every week I find myself charting the depths of a new technical problem that I hadn't previously even considered.
3
14
159
Software & mathematics are both fundamentally about iteratively building better abstractions. The difference is that math is encoded as proofs, which require special skills to appreciate, while software is encoded as machine-runnable code, letting anyone enjoy the results.
10
74
4
403
Programming is the art of coaxing reluctant computers to do what you want most of the time.
6
19
2
195
Lessons learned from detecting & stopping hundreds of actors attempting to misuse GPT-3:
Deploying and studying the real-world use of language models helps us learn more about safety and misuse than research alone.
As we advance our safety and policy work, we're sharing some of our findings to help others do the same. openai.com/blog/language-mod…
1
8
60
Humans are great at coming up with a single solution that meets any given constraints, and thereafter coming up with reasons why any other proposed solution won’t work.
So when optimizing for a goal, set aspirational constraints, and then relax them until a solution is found.
7
12
1
108
One of the least-appreciated skills in programming is writing anti-frustrating error messages.
A good error message should make it self-evident (a) what the user did, (b) what acceptable inputs are, and (c) how to fix the problem.
Can determine love or hate for your library.
245
663
50
8,979
GPT-3 for fixing OCR errors:
A rare & unknown gem of a paper — academia.edu/67825612/A_Soci… (from 1996!) analyzes the root cause of the AI winters.
The story you always hear is that the neural net people overhyped everything; the real story seems to be a dedicated smear campaign by eminent scientists.
12
34
5
330