OpenAI
-
I'm on the Microsoft Insider programme and the AI-aided Office is about to drop soon. Apparently it can search thru an email conversation and prepare a summary & timeline and print/post it to OneNote or Word in something like 20 seconds...
Absolute winner that.
-
@Victor-Meldrew The longed for future of never reading emails or answering phone calls again!
-
Another relevant take from today:
I'm convinced there is a certain class of people who gravitate to positions of power, like "moderators", (partisan) journalists, etc. Now, the ultimate moderator role has now been created, more powerful than moderating 1000 subreddits - the AI safety job who will control what AI "thinks"/says for "safety" reasons.
Pretty soon AI will be an expert at subtly steering you toward thinking/voting for whatever the "safety" experts want.It's probably convenient for them to have everyone focused on the fear of evil Skynet wiping out humanity, while everyone is distracted from the more likely scenario of people with an agenda controlling the advice given to you by your super intelligent assistant.
Because of X, we need to invade this country. Because of Y, we need to pass all these terrible laws limiting freedom. Because of Z, we need to make sure AI is "safe".
For this reason, I view "safe" AIs as more dangerous than "unsafe" ones.
-
@Victor-Meldrew I like the idea that the future of AI is all about accurate subtitles for 20th century films.
-
-
The only times I've thought that maybe I would turn to an AI (current generation) which might be able to help me with writing out some SQL or grep or regex... has been when I've also thought "I've got no fucking idea how I would even express what I want/need in english"
-
@Victor-Meldrew I like the idea that the future of AI is all about accurate subtitles for 20th century films.
Nah, the future is that this sort of capability can be produced in a few hours or days with AI and costs bugger all to develop.
-
@Victor-Meldrew said in OpenAI:
@Victor-Meldrew I like the idea that the future of AI is all about accurate subtitles for 20th century films.
Nah, the future is that this sort of capability can be produced in a few hours or days with AI and costs bugger all to develop.
that's one of the key issues is that the 'moat' for AI seems to not exist. PEople are doing crazy things with desktop computers ... it's good for us colelctively, but shit if you're trying to make money on it.
And then I hear that Microsoft are losting shitloads on their AI - burning computation power. So who knows
-
@Victor-Meldrew said in OpenAI:
@Victor-Meldrew I like the idea that the future of AI is all about accurate subtitles for 20th century films.
Nah, the future is that this sort of capability can be produced in a few hours or days with AI and costs bugger all to develop.
that's one of the key issues is that the 'moat' for AI seems to not exist. PEople are doing crazy things with desktop computers ... it's good for us colelctively, but shit if you're trying to make money on it.
Here's another thing. I use a Remarkable 2 and the API changed this year to a proprietary one and meant the Github OneNote integration plugin no longer works. Now there's repos using AI to reverse-engineer the API....
The implications are going to be big
-
@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