dexter89_kp3

Hi Peter,

Would love your thoughts on the recent EO on AI and it's possible impact on immigration.

this is awesome

this is awesome

Figma could have been much bigger than adobe in 5-10 years. Why not go public? Is the economic scenario so bad that they could not raise money?

The parts of the model and how to go about training it are well known (i.e VIT VQGAN, auto-regressive transformers etc). They also use LAION-400 for training which is an open source dataset. Can be replicated, but will take time, patience and compute.

This is not true for some of the other recent headline papers: DALL-E2, PALM, Imagen etc. datasets are the primary deterrent, model details are well known.

Can be done yes. Is it the best dataset to do what you describe, no.

You could train the initial model in sim and do sim2real finetuning. Someone must be working on this already

Self-driving car companies are already working on this. Check out: https://waymo.com/research/block-nerf/

NERF is a very active research area, and the progress from 2020 to now has been nothing sort of astonishing. In 5 years, I expect there to be fully generative NERF's in research i.e describe a scene, and a NN produces a full 3d scene that can you interact with.

There are on chain NFTs like cryptopunks. The rest of articles details around API centralization stand true.

"Email provides light-speed delivery, dirt-cheap cost, and brain-dead convenience on top of also being able to mail. So email actually never competed with mail"

These advantages are obvious now. See: https://www.youtube.com/watch?v=gipL_CEw-fk . My point being that blockchain does offer some new capabilities that were hard to build before. We maybe oblivious in what applications that may unlock.

Regarding money transfer, I know the technology exists, but all providers don't built it. That is native technology in blockchain.

Depends on whether the NFT is on-chain or off-chain. Both are valid.

I don't understand what problem email can solve that mail cannot.

Sorry to be sarcastic.

I am still skeptical of most web3 use cases touted on twitter, but have personally experienced a few that have made me pause and be a bit more open minded.

Personal experiences: - Transferring money was in seconds, compared to Fidelity/Venmo which took 3 days!!! - Very easy to loan money against crypto holdings vs stocks - ResearchDao's like ResearchHub

LinkedIn is basically a giant water-cooler where everyone can show-off. That i s the incentive, the network decided to promote.

Interesting. My current side project is exploring semantic search using sentence transformers. Will definitely check this out.

With increasing evidence that self-supervision works for multiple tasks and architectures, and sim2real/simulated data being used in industry I do not see this as an important concern.

Framework laptop is not possible without the supply chain investments made by large conglomerates.

Not in favor of the headline. The deepmind blog clearly states "Our results suggest that ML can complement maths research to guide intuition about a problem by detecting the existence of hypothesised patterns with supervised learning and giving insight into these patterns with attribution techniques from machine learning"

I invested some token money on OHM to learn more about it. The APY figures will come down in 1-2 months based on certain funding/staking levels. I stayed away from all forks of OHM.

Lot of interesting projects happening in DeFI 2.0 and DAO space. One example: https://www.klimadao.finance/

So... the failure may not be due to ML, but due to people :) See this blind thread: https://www.teamblind.com/post/Did-Zillow-fail-miserably-due...

Revisit your writing fresh, and read it as if you were reading someone else's writing

This looks awesome!

Could you give an approximate cost of fine-tuning a model like Bert or even GAN given this system?

Just want to get a sense of cost of using the system.

Decisions taken by British still haunt India to this date. Especially their division of states.

Small government != efficient government.

A government can spend billions of dollars efficiently.

As an example look at current COVID rapid testing kits. The US government is purchasing them at 4x the cost in other developed economies. California gave away $31 billion dollars in unemployment benefits to criminals.

This is what frustrates me. I am happy paying 40% tax as long as I see it making a difference in society.

I wonder why there is so much thought and work on raising taxes but not on efficient spending by the government. There are several examples around where government spending is inefficients and sometimes corrupt.

Any way to stir to conversation and political thinking on efficient spending?

Gamifying my exercise and workouts, also making it more social.

I find this complaint surprising. You can turn notifications off. I have everything disabled apart from health metrics on my watch.

I gave the timestamp where they start talking about the problem they are trying to solve using transformers. As you said it is around 1hr mark

They do. Karpathy spoke about it in on Tesla AI day. They use it for transforming image space to a vector space.

See: https://youtu.be/j0z4FweCy4M (timestamp 54.40 onwards)

Kudos to the Waymo team.

I was a self driving skeptic 2-3 years back. Now given the advances in both hardware and NNs, I do see possible solutions in the next decade.

Was not defending this measure, my question was out of curiosity