dougmwne

27 ms·

331.
by dougmwne 3y ago
The ability to write a polite and on-topic email used to be a valuable skill to demonstrate. Not so much anymore.
332.
by dougmwne 3y ago
ChatGPT 3.5 is likely much smaller than GPT-3’s 175b parameters. Based on the API pricing, I believe 8k context GPT-4 is larger than 175b parameters, but less than 1t. https://openai.com/pricing
333.
by dougmwne 3y ago
So far Bard has been very behind GPT-4 and very prone to hallucinations. It feels much more like talking to an autocomplete keyboard than a baby-AGI like GPT-4. Hopefully this PaLM 2 model closes the gap, because the gap in a large.
334.
by dougmwne 3y ago
This is a bunch of hallucinated gibberish.
335.
by dougmwne 3y ago
This is just “ Minnesota nice”. Cultural expectations of politeness are not really correlated with any specific ethics. As to the downvotes, you are inadvertently opening a can of political worms around competing values of the coasts and th
336.
by dougmwne 3y ago
It seems quite clear that we are just a few years from being able to wire up less than 20k of nvidia hardware and produce a GPT-4 class model. All the methods and datasets are out there. Plenty of pre-trained base model weights too. The hor
337.
by dougmwne 3y ago
Especially considering that it is so widespread in nearly every creature with a brain. And it’s not simply a period of motionless energy conservation but has very specific neural patterns. The science is definitely zeroing in on a connectio
338.
by dougmwne 3y ago
Definition #2 clearly refers to and includes an industrial chemical process as you can see in the usage examples and synonyms. This is how language works. One word can have many meanings and shades of meanings. When people talk about chemic
339.
by dougmwne 3y ago
Extremely expensive, though as I understand it, the goal is to get maximum performance out of a given model size so that you can actually inference at product scale. A few extra million for training is expensive, but then consider what it c
340.
by dougmwne 3y ago
Once you prune your model, can you get even better performance by re-training it? I've heard theories that this is the function of sleep in brains.
341.
by dougmwne 3y ago
You can’t redefine words as you see fit to suit a simplified model of the world. Meaning is complex and shaded by situation and intent.
342.
by dougmwne 3y ago
I didn’t mean to call all GPT-4 output gibberish, just this commercial script. But now that you mention it, if a person started speaking unconsciously with no meaning connected to their words and no intent behind their communications, we wo
343.
by dougmwne 3y ago
It is possible for a statement to be technically correct, but practically incorrect. Words have different definitions and meanings in different contexts. If you try to force a meaning from one context into another context, you have made an
344.
by dougmwne 3y ago
Here’s the original creator of that pizza commercial explaining the workflow. GPT-4 was used for the script, which sounds plenty plausible to me considering that it’s gibberish. After Effects was used for the editing, transitions and graphi
345.
by dougmwne 3y ago
It’s shadow IT. You are not supposed to leak confidential company data to any other company unless you have the appropriate vendor agreements in place. It’s like loading the source code to Dropbox when you are supposed to use bitbucket.
346.
by dougmwne 3y ago
Seems like a no lose bet no matter the odds. If you win, life goes on and you can collect a bit of money. If you lose, money becomes worthless and the surface of the earth gets transformed into a computing substrate.
347.
by dougmwne 3y ago
And that will happen no matter what. Polls and prognostications will always exist and probably have always existed back to the first tribal chieftain election. It’s just that instead of getting totally inaccurate predictions that only serve
348.
by dougmwne 3y ago
If anything, I suspect this detailed polling of making elections closer. Millions of voters aren’t reading 538. It’s a distinctly wonky thing to do. But people really into politics, the kind of people who volunteer for campaigns and make la
349.
by dougmwne 3y ago
I really think it has a lot to do with the situation and expectations. And probably you as well. I think it’s completely plausible to deliver expected value in 7 hours, though you are probably looking at either a very meeting heavy role or
350.
by dougmwne 3y ago
Why would Facebook need to raise all this cash? Don’t they have plenty of profit? Aren’t interest rates high enough to discourage borrowing against future profits like this?
351.
by dougmwne 4y ago
What a charming celebration of failure. We need more of this in our businesses and our lives. It’s a good way to give people permission to aim high, take big risks and help pick up the pieces when it doesn’t work out. I love the gift card f
352.
by dougmwne 4y ago
I’m sure that’s the case and they were also widely reported to be selling each unit at a loss for 299.
353.
by dougmwne 4y ago
Yeah, I’ve been reading they are refocusing on gaming. It’s a smaller market, but something they have awesome product market fit on. Metaverse/Social still seems plausible to me, but maybe in 5-10 years when we have photorealistic avat
354.
by dougmwne 4y ago
It’s at about 20 million now. I’ve seen estimates that about 1/3 are MAU.
355.
by dougmwne 4y ago
Quest 2 is actually up in price from 299 to 399. It’s not currently on sale. By all indications, it wildly exceeded initial sales expectations. Quest 1 is estimated to have sold 1 million units and Quest 2 did an estimated 20 million units,
356.
by dougmwne 4y ago
That seems expected. They are very late in the cycle of the current Quest 2 headset and there have been no major game releases for months. Quest 3 with a new stable of launch titles is likely less than 6 months away. Other moves have been s
357.
by dougmwne 4y ago
If your savings can be measured in decades, then you should look into some of the math and simulations behind the FIRE community. While considering yourself financially independent may be more of an academic exercise, the most important thi
358.
by dougmwne 4y ago
I’m not sure if I ever believed that hogwash. I think it was a fantasy meant to try to justify ludicrous stock prices on unprofitable companies. After 14 years of serious investment, it’s not clear if self driving cars are ever going to bec
359.
by dougmwne 4y ago
It's a good question, but also helpful to point out that one of the beauties of these models is that you can train them once and deploy to many use cases. The same model can be used by Github, Bing, Office 365, Azure and so on. And as
360.
by dougmwne 4y ago
Yes, it has absolutely nothing to do with the exact ideology and everything to do with a narrowing of what is publicly acceptable. For an example, compare any hollywood comedy from 10 years ago to one from today. So it's neo-victorian

More ›