All of platers's Comments + Replies

These are very impressive! It looks like it gets the concepts, but lacks global coherency. 

Could anyone comment on how far we are from results of similar quality as the training set? Can we expect better results just by scaling up the generator or CLIP?

Using CLIP is a pretty weird way to go. It's like using a CNN classifier to generate images: it can be done, but like a dog walking, we're more surprised to see it work at all.

If you think about how a contrastive loss works, it's perhaps less surprising why CLIP-guided images look the way they do, and do things like try to repeat an object many time: if you have a prompt like "Mickey Mouse", what could be even more Mickey-Mouse-y than Mickey Mouse tiled a dozen times? That surely maximizes its embedding encoding 'Mickey Mouse', and its distance from non-Di... (read more)

There is a paper describing the architecture

It looks like the system is comprised of many independent skills and an algorithm to pick which skill to use at each state of the conversation. Some of the skills use neural nets, like a CNN for parsing images and a RNN for completing sentences but the models look relatively small. 

Wouldn't have expected to read this in the abstract of an AI paper yet. Also this feels kinda creepy as a caption.

Could you elaborate on step 4? How can you buy 10 shares of each bucket if you only have $10? Isn't the total cost 14.80 * 10?

5Scott Garrabrant2y
Predictit promises that only one will be yes and immediately gives you back the liquid money when you arbitrage.
[I'm not your financial advisor. I don't know your financial situation, and this is not financial advice. Leverage is a double-edged sword. Don't Bet the Farm. []] Depends on what you need it for. 3x ETFs are probably plenty for a basic risk premium portfolio. LEAPS might have more variety, but you have to buy 100 share increments (which makes it harder to balance in a small account), and you have to deal with vega risk and occasional rolling hassle and costs. Portfolio margin [] is probably the best way overall. But I haven't seen it available for less than a $100,000 account. You can get almost 2x on an ordinary Reg-T margin account though. Brokers may charge outrageous amounts of interest for margin loans. You can sort of work around this by financing with box spreads. Outside the US, you can probably get 5x on CFDs. I can't trade them, so I haven't looked into them too much. Leverage up to 20x is easy to get for Forex (and maybe 50x for the major pairs), because Forex volatility is so low, you need a lot of leverage to bother with it. But the market is mature and efficient enough that I don't recommend trading it without some automation. Grinding out an edge that small is really tedious if you have to do it manually.

There is a very extensive discussion of a UPRO/TMF strategy here. One thing to note is taxes severely decrease the returns of strategies which require frequent re-balancing.

Taxes are going to be a problem for any active strategy. "DON'T PANIC" from part 2 is mainly a warning about trading too often. In the U.S. at least, you can avoid some taxes by trading in a Roth IRA, but it's problematic for early retirement. There are limits to how much you can contribute. If you mess up and bet the farm that limits how quickly it can recover. You are allowed to withdraw your contributions early without penalty (but not your returns, with a few exceptions). You also can't trade on margin, but can use ETFs with leverage.

Have you rechecked the data recently?

I did, but I’d didn’t bother crossposting here since the first one didn’t get any engagement/karma: []

I see, thanks for clarifying!

Why do you think "lesser" AI being transformative is more worrying than AGI? This scenario seems similar to past technological progress.

6G Gordon Worley III2y
I didn't say GPT-N is more worrying than AGI, I'm saying I'm surprised we near term have to worry or be concerned about GPT-N in a way I (and I think many others) expected only to have to worry about things we would all agree were AGI.