I think it can. I've been working on getting an LLM (Claude Sonnet 3.7) to add missing type annotations to python code. When I naively ask it "please add types" then like you said it has about a 60% success rate and 40% hallucination rate as measured by "would an expert human have come up with the same type annotations and did they pass the typechecker".

But when I have a much more careful use of the LLM, micromanaging what sub-tasks it does, then it has a 70% success rate, and 30% rate of declining because it didn't have confidence to come up with an answer. Effectively there were no more hallucinations. (I got these numbers by spot-checking 200 cases).

So I think hallucination can be solved for some tasks, by the right kind of task-specific micromanagement and feedback loops.

OpenAI has $12bn annual revenue, about 3% that of Apple, about $3million per employees per year (compared to $2 million per employee per year at Apple).

I think OpenAI has a huge amount of growth potential even just from predictable growth over the next several years, even if steep changes towards AGI don't come.

Like I said, I figured 8k would be enough resolution for soccer. As for 16k, I imagine that something with bandwidth for 16k would translate that bandwidth into twice the frequency for 8k, which would be ideal for soccer.

Do you watch soccer? 4k resolution means a player's head is about 14 pixels high, not enough to make out much beyond a blob of color; their jersey is 60 pixels high, enough to make out the number but not much more. Doubling the vertical resolution (i.e. going to 8k) would likely be enough to let you make out similar detail to what you'd see in real life. (Frame rate is another issue: HDMI 2.0 allows 4k at 60hz which is too slow when panning in a soccer game; HDMI 2.1 allows 4k at 120hz which is probably enough). I think that 16k is probably the right bandwidth to get soccer looking good.

Do you do VR? 4k per eye isn't good enough for VR yet. It's possible that 16k will be, but we might still need more.

Do you watch the gorgeous film classics like Lawrence of Arabia? One of the (many) things that make it look great is that it was shot on 65mm, equivalent to about 12k resolution.

The military team scrub *alerts about objects *.

Three days later, the full unscrubbed image is shared with astronomers.

I guess they don't and can't care about stable orbits, but they want to retain surprise for dynamic situations.

It's easy to regulate AI art the state level.

"Any job offer for a job based in California must adhere to the following AI disclosure".

"Any mortgage offered in a Californian property must satisfy the following AI disclosure"

etc.

AI regulation need not be about regulating AI innovation; it's enough merely to make sure it's applied fairly. And almost all real-world applications are indeed local.

I used to work at Microsoft. My employment contract specifically called out a load of personal pre-existing projects, plus ongoing and future ones, and stipulated that MS would have no ownership nor claim. I did ask for these callouts, but they were happy to go along.

I'm a software developer. Part of AI is like if I had 200 interns working for me -- some of them smarter than me and already more knowledgeable about some areas, some of them not, none of them familiar with my team's codebase. There are real cases where I could get those 200 interns to do real useful work and would want to! e.g. if I create a very detailed playbook of how to make certain code improvements, ones that wouldn't be worth my time to do myself one-by-one, but if I had 200 interns and an automated way to verify that they did a good job, then sure!

The article says "manage a team of AI agents". Managing in this sense isn't like managing a human; it's like writing a shell-script to manage some bulk process.

I think there is for sports. Watch soccer on a 4k TV. The camera is usually pulled back far enough to see a lot of the field, so each individual player on a 4k screen (3840x2160) is about 150 pixels tall, and the number of their jersey is about 30 pixels tall. That's usually not enough for me to make out what's happening. I can make it out better live in person. An 8k screen I think would be enough to make it out. I'd sit closer to it than your 8' if I wanted to watch. (Likewise, at IMAX I like to sit about 5 rows from the front so the screen fills my peripheral vision).

I'm a senior developer and use LLM assistance multiple times an hour. >90% of the time I find something valuable in what it produced (even though rarely accept what it suggested directly, and I often write or rewrite every single line).

So what value does it have if I'm not accepting its code overall? Lots of value....
1. As a professional I produce code that (1) I can reason about why it's correct in all possible environments, (2) I'm confident that the way I've expressed it is the best it can be expressed in this situation. The LLM can spit out several different ways of expressing it, helping me assess the landscape of possible expressions, allowing me to refine my evaluation of what's best. (It doesn't yet help at all with reasoning about correctness).
2. About 10% of the time I accept some of the lines it suggested because they save some inescapable boilerplate. Or it spits out enough boilerplate to tell me hey, I need to invent an abstraction to avoid boilerplate here. I'd have gotten there myself too, just slower.
3. Sometimes I find myself learning new idioms or library functions from its suggestions.

I think management is right to be AI crazy. LLMs have increased the rate at which I solve business needs with high quality code, and I think my experience generalizes to other people who are willing to take it on and "hold it right". (Of course, there'll be vastly more people who use it to write low quality code faster, and it'll be up to management to separate the good from the bad just like it always has been.)

The way I use it: I write a method stub with just signature and docstring, or a class stub, then ask the LLM to flesh it out.

Do I ever use what the LLM produced? -- never; I always discard every single line it produced, and supply my own.

Do I ever benefit from what the LLM produced? -- usually yes, about 90% of the time. It shows me fragments or idioms or API usages that maybe I didn't know about, or it shows me a cumbersome approach which tells me where to focus my efforts on finding something more elegant. Often I'll copy/paste some expressions or lines from it into my solution. About half the time I follow up with further prompts, e.g. "The code you wrote ; please rewrite it to ."

When I'm writing code, I'm kind of proud of my professional skill, and for every single line I produce I asked myself (1) can I prove that this line is correct? (2) am I confident that this line/method/class/file/subsystem is the optimal way to achieve what should be done? Having the LLM spit out its (non-optimal) solutions helps me assess the design landscape, is a cheap way to see other ways to achieve what should be done and hence improve my judgment on whether mine is optimal.

I have a Dell P322QE with native resolution 3840x2160, i.e. 4k. But MacOS still doesn't turn on proper scaling for it.

> System Settings -> Displays -> Larger Text vs. More Space. It changes scaling, not the resolution.

What you describe changes the RESOLUTION (at least it does on my mac connected to my Dell P3222QE). Indeed when you hover over one of the icons in "larger text vs more space", a little hover text shows the resolution that it's going to pick.

And if you click "Displays > Advanced > show resolutions as list" then it replaces those "Larger text vs more space" options with a dropdown of available resolutions.

> What on Earth are you on about? MacOS has better scaling than any other OS out there.

I don't think so? I'm 50yo so my eyesight is getting worse. I need large text to be able to read it. My monitor resolution is 3840x2160.

When I used to use Windows, to get everything in large text (menus, dialogs, prompts, ...) I'd stay in 3840x2160 resolution and choose "large fonts 150% or 200%". The UI adjusts: I get the large fonts I need, but staying at native resolution, so all the lines+curves in the fonts are clear and crisp, and photos have all their detail.

On Mac to get everything in large text, the only option I have is to bump down the resolution, currently 3008x1692 but I'll probably have to go to 2560x1440. This gives me the large fonts I need. But it's at a lower resolution, so the fonts look pixelated, and pictures can't be displayed in as much detail.

Did I understand you wrong? Is there some other way to get MacOS to have nice scaling? I haven't found it, but I'd dearly love to.

You can't manage an AI. That doesn't make sense. It's like managing a hamster or a dolphin or a horse. The only thing you every manage is the *humans* who wrangle the AI/hamster/horse.