It's not a shitty analogy. Most people completely understand what it's getting at. Inability to understand is a reflection on you, not the analogy. And being pedantic is not a positive trait or intellectually impressive.
This has been my experience as well. I cannot understand how people are saying it has given them order of magnitude speed improvements in shipping.
Most code on github is by definition average, which is usually a few levels below the quality of an expert programmer. So it's not surprising the output of code LLMs is poor to average quality. For any larger block of code it outputs, a non-trivial amount of editing is required if a higher bar of quality is wanted.
Does anyone get an unpleasant feeling from these typical "founder" types? Something always just feels deeply off about them. They all seem like carbon copies of each other too.
Changing the fee from 3%-5% total to 12.9% (17% on paypal!) total when you already have a large customer base there, meaning you'd have to painfully migrate them to another provider, is not a nice thing to do in my eyes.
And also why I'm working on completely migrating off.
Just here to post this is another reason to get off Gumroad when you get a chance. Do NOT use Gumroad. Using Stripe directly will save you so much money. Still bitter at those exploitive jerks jacking the payments price to a total of 12.9% (and something like 17% when through PayPal).
I was always surprised at how AMD hasn't already thrown a bunch of money at this problem. Maybe they have and are just incompetent in this area.
My prediction is AMD is already working on this internally, except more oriented around PyTorch not Hotz's Tinygrad, which I doubt will get much traction.
Sam Altman said it himself. He seems like a reasonable source.
If you're familiar with other fields of AI, adding more and more layers to ResNet was the hotness for awhile, but the trick stopped working after awhile.
As mentioned, much of our discussion is rote parroting. I can usually go into any hackernews thread and roughly know what the top discussions are going to be. It's not surprising that an AI trained on a large portion of the internet would thus look human like.
If you really poke at GPT, you begin to realize it's fairly shallow. Human intelligence is like a deep well or pond, where as GPT is a vast but shallow ocean.
Making that ocean deeper is not a trivial problem that we can just throw more compute or data at. We've pretty much tapped out that depth with GPT4 and are going to need better designs.
This could only take half a decade or it could be half a century. Plenty of enterprises stagnate for decades.
>Using the current approach, we can create AIs with the ability to do any task at the level of the best humans — and some tasks much better. Achieving this requires training on large amounts of high-quality data.
Wrong. Full self driving is still not here despite access to a huge amount of high quality data.
>I get the cold sweats when I see methods longer than a few lines of code
This guys code might not be as readable as he thinks. I do not want to read through your hundreds of 10 line functions because a few people said "long functions bad".
I liked the portion on naming. My God the amount of AbstractFactorySingletonViewModel garbage I see is astounding.