OpenAI
-
@Kirwan I've been using chatgpt for writing/rewriting and tidying up SQL. Any idea if that's the best source or is there something like copilot for plugging into an appropriate app?
Copilot runs on GPT4, so either of those. I use VSCode, and the integration copilot has with that is awesome.
Pretty much can get you 80% of crappy tasks like documentation or tests. Great at working out what someone else’s code is doing.
I’ve been using it suggest alternative solutions as sort of a training tool. See if there more efficient ways to skin the cat.
Good at picking up missing logic.
It does have plenty of flaws. It’s confidently incorrect about 20% of the time. You have to practice the prompting and hand hold a bit to start. Once your work out how much detail it needs to be useful then it gets better.
The IDE integration is spooky sometimes. The suggestions can build full functions that are exactly what you were intending. Works better if you have tabs open with the code in the supporting areas, and point it at line numbers, etc
I refactored some broken old code today. I figure I got through a solid two days work in four hours.
-
You can also see where this is heading towards the Star Trek computer where you talk or type a natural language question to complete a task.
“Where is the setting to do X”
“Make a spreadsheet with these columns form the data in this folder”
Windows copilot’s going to transform the interface to how we work with computers
-
This Q* sounds interesting. They have it teenager level of maths using reinforcement learning. If they have that actually working, then it can get smarter.
The big rumour is Sam telling the board about it caused the panic.
What’s the Chinese curse? May you live interesting times.
-
-
-
This was a good listen on AI
-
I wonder if they are restricting prompt response length due to load?
I did consider that but can't comment because I'm on the wait list, having spent too much time finding the limitations of its predecessor. And that, I'm certain has capacity constraints which are totally understandable.
-
I should imagine those problems will rapidly disappear now AWS is targeting business AI
-
Many AI Safety Orgs Have Tried to Criminalize Currently-Existing Open-
<p>I've seen a few conversations where someone says something like this:</p> <blockquote> <p>I've been using an open-source LLM lately -- I'm a huge fan of not depending on OpenAI, Anthropic, or Google. But I'm really sad that the AI safety groups are trying to ban the kind of open-source LLM that...
-
The power of this stuff continues to amaze me. For some reason a TV series I saw when I was about 12 came back to me. Thought it was a Francis Durbridge mystery and could only remember a character called Tolley.
A good test for CoPilot I though - found it after I asked 3 questions...
Babbage: Fei-Fei Li on how to really think about the future of AI
