HNHacker News
TopNewBestAskShowJobs

waynenilsen

535 karma · joined January 14, 2015

submissionscomments
waynenilsen··on Generative AI set to affect 300M jobs across major economies
While negative impacts are specific and large on an individual level, positive impacts are broad and diffuse but vast. This has always been the issue with various kinds of protectionism. Change hurts. We will all be better off. We should consider how to help these people also.
waynenilsen··on Twitter's Recommendation Algorithm
Shadowbanning was real and widely applied. That is the human part of the algorithm (manual mode) and it was very politically skewed
waynenilsen··on RFdiffusion: Diffusion model generates protein backbones
So I guess this means easier drug discovery? Honesty those wiggly diagrams are meaningless to me I have no bio background
waynenilsen··on BlenderGPT: Use commands in English to control Blender with OpenAI's GPT-4
I think real CSI "enhance" is coming also although i'm sure all kinds of details will be imagined and interpolated
waynenilsen··on BlenderGPT: Use commands in English to control Blender with OpenAI's GPT-4
Unfortunately the context window is too short. I have been wanting this also. I think there may be a place for microtraining whereby the model weights are re-trained with your new data. It can include everything in your organization.

In the long run I think that continuous training will be obvious with new models rolled out on a very high frequency if not continuously.

waynenilsen··on ChatGPT can now call Wolfram Alpha
I wonder if they're doing something like ReAct under the hood here.
waynenilsen··on Prompt Engineering: Steer a large pretrained language model to do what you want
LLMs will become another building block and we will once again stand on the shoulders of giants.
waynenilsen··on Prompt Engineering: Steer a large pretrained language model to do what you want
I have been looking for this! Chatgpt should have something similar with different flows for known good prompts. Templating for best practices for prompts seems pretty obvious but is very underexplored. An empty text box is almost the definition of choice overload.
waynenilsen··on Show HN: GPT Repo Loader – load entire code repos into GPT prompts
Software engineering will never be the same. The LLM will teach its users about programming. This is the worst LLM tech will ever be. That is an incredible statement. The rate of error will decrease to near zero or at the very least significantly better than human. Universities will resist at first but the new tools that will emerge will be core curriculum at university. Just as I now don't use the pumping lemma day to day, programmers of the future will not write code. They will primarily review and eventually AI systems will adversarially review and programmers will do final review.

All programmers will become translators from product vision to architecture implementation via guided code review. Eventually this gap will also be closed. Product will say: Make a website that aggregates powerlifting meet dates and keeps them up to date. Deploy it. Use my card on file. Don't spend more than $100/month. The AI will execute the plan.

Programmers will come in when product can't figure out what's wrong with the system.

waynenilsen··on Show HN: GPT Repo Loader – load entire code repos into GPT prompts
Fully automated junior swe but on hyper speed. Natural next step
waynenilsen··on GPT-4
This is the worst it will ever be, period
waynenilsen··on A study on the extent of global air pollution
Seems like too few categories, safe or not safe? Clearly this is a spectrum
waynenilsen··on Pre-seed, seed startups with employees in office have higher revenue growth
There is a very clear statistical standard here "difference of means" test at the very least. Very very small dataset. Lots of options all of which would most likely conclude no statistical difference and low power.
waynenilsen··on ChatGPT: The Slickest Con Artist of All Time
We are completely off topic at this point but it is worth mentioning that due to the pseudoanonymous nature of bitcoin and ethereum, they are actually quite bad for those things and tokens such an Monero, ZCash, Grin, Beam, Mina and many others are much better suited to illegal transfers.

Also worth mentioning, what is illegal strongly depends on jurisdiction and can sometimes be moral despite their legal status.

waynenilsen··on Cargo airships could be big
It seems about 2x faster plus overland capabilities should be great for some use cases that currently depend on canals.
waynenilsen··on Orion Browser
I like that I am the customer. Now they just need go make a twitter clone for $5/mo
waynenilsen··on 1min high-intensity exercise 3x a week improves fitness as much as 3x aerobics (2016)
I would also add that hit has a better chance of improving overall metabolic health raising resting metabolic rate and allowing you to eat more in the long run. Short run, low and slow will increase total daily expected expenditure but hit is a better long term investment.
waynenilsen··on VALL-E: Microsoft’s new zero-shot text-to-speech model
Perhaps digital signatures will help with authenticity issues. It may become necessary. Transcoding presents an issue I wonder how it may be solved
waynenilsen··on Tell HN: The regret of parenting and how to handle it
Do you have kids?
waynenilsen··on US National Debt Tops $31T
There is a mathematical certainty implied here that I believe many people here can understand. Compounding interest is exponential. We are borrowing money to pay the interest, this will continue until something breaks.
waynenilsen··on US National Debt Tops $31T
incorrect incentive structure
waynenilsen··on New Text and Array Functions for Excel
I am very surprised they have not yet embraced the dplyr/tidyr-style melt/cast gather/spread pivot/unpivot functions for arrays
waynenilsen··on IPFS and Their Gateways
This article made a false statement

> the gateway is in full control and can inspect and tamper with the data as much as it likes. And there is no way for the client to know or detect if it is happening.

Regarding data tampering;

IPFS uses content based addressing, the point is that the client _can_ know by computing a message digest of the incoming data. The verification is fast and cheap.

waynenilsen··on Dusting “Attack” via Tornado Cash to Public Wallets
It is actually worse than you think, the entire account ends up having "interacted" with a "sanctioned" entity :( account owners may be subject 10y in jail if any prosecutor would bring a case. This is true for any tokens and NFTs associated with the account as well as the ETH.
waynenilsen··on Ask HN: How is the remote vs in-person trend looking?
i am never going back to the office
waynenilsen··on Htmlq: like jq, but for html
part of the problem with this is that HTML is mostly not valid XML
waynenilsen··on El Salvador makes Bitcoin legal tender
One vision is that checking accounts are custodial LN solutions with savings accounts as cold storage.

Discussion with the "Bitcoin Beach" people https://stephanlivera.com/episode/279/

waynenilsen··on Vapor – Decentralized Web over Bitcoinized HTTP
I found it

> The protocol serializes an HTTP request and wraps it inside a Bitcoin OP_RETURN output script. This is then signed with a Bitcoin wallet's identity private key and included as another output script in the transaction. The resulting Bitcoin transaction is sent to a destination Vapor API endpoint over HTTP.

> A Vapor endpoint only accepts HTTP POST requests with a raw bitcoin transaction as payload. The endpoint parses the transaction to extract out authentication information, validates the signature, and then forwards the extracted HTTP request to relevant API endpoints.

I actually think this is probably better suited as a library, then there is no need for any intermediary calls that will inevitably increase latency. Also, would the vapor server somehow send the information about the public key on to the destination server?

waynenilsen··on Vapor – Decentralized Web over Bitcoinized HTTP
Interested to see the code, I still don't totally understand how this works. Is there a generic vapor server that re-routes http requests to the hosts specified in the op-return? If so, what does that give you? If not, then the caller needs to submit this bitcoin transaction as a ... payload to some app-controled proxy that might inject some associated account information by querying a db then route the call to the right endpoint with that info in trusted headers or something?

I think I understand the concept as something like "login with your bitcoin identity" but I'm struggling to understand practically how what you're envisioning is meant to work.

I actually think this could be made far more generic as "login with public key identity" for any crypto scheme, Ethereum or just about anything that is not mimblewimble and/or hyper privacy focused should work similarly.

waynenilsen··on The Long-Term Stock Exchange Opens for Business
Regarding "a level playing field for all market participants" I don't believe that in an exchange that supports HFT there can be a truly level playing field. There are some technologies that can help mitigate the effects of HFT on the market. One is the approach that AZX [1] took where you do large and infrequent dual auctions. Another approach is to go with an Automated Market Maker (AMM) technology and eliminate the need for order books all together as has become popular recently in the cryptocurrency space. Some background research can be found here [2] but what really took off was Uniswap's innovative use of a "Constant Function Market Maker" [3]. This allows average individuals to provide liquidity while getting the benefits of what is normally only reserved for high frequency traders. I believe that an institutional grade AMM program would be greatly democratizing for the industry.

[1] https://en.wikipedia.org/wiki/Arizona_Stock_Exchange [2] http://reports-archive.adm.cs.cmu.edu/anon/2012/CMU-CS-12-12... [3] https://medium.com/bollinger-investment-group/constant-funct...

← PreviousPage 4 of 7Next →