HNHacker News
TopNewBestAskShowJobs

iandanforth

11,490 karma · joined December 18, 2009

ml/ai/rl - neuroscience - robots - growprammer

@iandanforth

Source available - https://www.openhumans.org/member/iandanforth/

meet.hn/city/us-Pittsburgh

submissionscomments
iandanforth··on Show HN: Never Apply Again
So you'll be an agent? I think this is a great idea. My major concern is this, " you'll pay me whatever you want when you accept an offer." There's nothing worse for a business relationship than unspecified financial terms. It will be bad for you and your clients. Pick a number, a percentage, something, and put it in writing up front.
iandanforth··on Ask HN: Quiet Quitting as Tech Lead
Been there, regret that. Your time is precious, don't waste it on companies like this. Unless you can't quit, do so immediately and find anything else to do with your time. Of course if you're really just using your paid time to work for different work, no worries, that's a decent strategy as well.
iandanforth··on Show HN: Open-source macOS AI copilot using vision and voice
For example, one of these:

https://opencompass.org.cn/leaderboard-multimodal

iandanforth··on Show HN: I Remade the Fake Google Gemini Demo, Except Using GPT-4 and It's Real
Looks like, again, this doesn't have GPT-4 processing video as much as a stack of video frames, concatenated and sent as a single image. But much closer to real!
iandanforth··on FDA approves a CRISPR-based medicine for treatment of sickle cell disease
You're going to want to avoid using language straight out of GATTACA if you want to have a real conversation on this topic.
iandanforth··on Fairphone 5: Keeping it 10/10?
I'd buy one if it had a headphone jack!
iandanforth··on Gemini AI
I'm curious how it performs on the abstraction and reasoning challenge!
iandanforth··on Sequential modeling enables scalable learning for large vision models
If this paper is coming out of BAIR at a max 3B parameter model I suspect we'll quickly see much larger models from the industrial players. Hopefully Mistral takes an interest and releases an OSI licensed model.
iandanforth··on Sequential modeling enables scalable learning for large vision models
You might enjoy 'On Intelligence' by Jeff Hawkins to learn more about this hypothesis. (It's an older book / theory at this point but still worth reading IMHO)
iandanforth··on 23andMe hackers accessed a whole lot of personal data
I recommend you only send your genetic data to companies if you're willing to have it exposed. Then again I also recommend you make your genetic data public, for the common good. Here's Mine: https://www.openhumans.org/member/iandanforth/
iandanforth··on Download all of Wikipedia on your phone
Can this be done for a Kindle?
iandanforth··on Why Isn't Landfill Mining More Popular?
I'd love to see 'mining' defined more broadly here. Sure metals are a great target but every useful and valuable material known to man exists in landfills. A "full cycle" or "total recovery" mindset would be far more exciting. Treat everything as potentially valuable and you get better economics. I'd love to join a company who had this as their mission. Sure, start with metal recovery but continually build out new recovery processes until you can "ingest" everything and turn it into something useful.
iandanforth··on We have reached an agreement in principle for Sam to return to OpenAI as CEO
I apologize, the comment's irony overwhelmed my snark containment system.
iandanforth··on We have reached an agreement in principle for Sam to return to OpenAI as CEO
"In practice, board bylaws and common sense mean that individuals ... don't do stupid shit."

Were you watching a different show than the rest of us?

iandanforth··on Claude 2.1
I've been having fairly good success with Claude 2 via AWS Bedrock. So far I haven't needed to use the full context window of the existing model, but some of my future usecases may. I look forward to testing this model out if/when it becomes available in Bedrock as well.
iandanforth··on Orca 2: Teaching Small Language Models How to Reason
Released under the MS Research License, so not OSI and non-commercial, for the curious.

https://huggingface.co/microsoft/Orca-2-13b/blob/main/LICENS...

iandanforth··on OpenAI staff threaten to quit unless board resigns
I'm thinking more like "24"
iandanforth··on Sam Altman, Greg Brockman and others to join Microsoft
If Ilya is there many will. If Karpathy stays many more. If Alec Radford stays then ...
iandanforth··on Sam Altman, Greg Brockman and others to join Microsoft
Seems like it will create a Deepmind/Google Brain style split within MS.

MSR leadership is probably a little shaken at the moment.

iandanforth··on OpenAI's board has fired Sam Altman
The statement claims he is no longer board chair but will stay with the company. Do you have other info?
iandanforth··on OpenAI's board has fired Sam Altman
As noted in the release: "The majority of the board is independent, and the independent directors do not hold equity in OpenAI."
iandanforth··on GitHub Actions Are a Problem
Other people must have very different setups to warrant this kind of additional overhead and tooling.

In our repo local testing is first. You check out the code and can run the tests. The github runners do the same steps the readme encourages humans to do when running tests locally.

Also I'm not sure what duplication the author is avoiding with their DSL as Github actions can be broken into components and reused and parameterized. Of course if you don't like YAML you're never going to be happy until you've wrapped it in your own layer of tech debt.

iandanforth··on Casino-like apps have drained people of millions
I am reminded of this Penny Arcade comic: https://www.penny-arcade.com/comic/2005/04/25/an-artifact-of...

"It's so pure I think I'm going to cry."

iandanforth··on Real-time image editing using latent consistency models
Yep! I saw YouTube videos of that workflow before posting the comment. I will say that I generally don't need Blender levels of control (or maybe I do and just don't know it yet) so a simplified set of primitives in a stripped down UI would be great.
iandanforth··on Starship’s second flight test
I recommend the above link because it gets even weirder than you might surmise from the parent's comment.
iandanforth··on Real-time image editing using latent consistency models
This is awesome. For future inputs please give me a 3d scene with camera controls into which I can drop primitive shapes, and human pose dolls so that I can frame shots like I would with a camera. I can't draw so even figuring out how to suggest "draw the image from the top left as the person looks over their shoulder" to the model is something I struggle with.
iandanforth··on NYPD barred from telling people they can't film in precinct stations
The line between 'annoying ego-trip' and 'act essential for the preservation of freedom and democracy' is surprisingly thin. In many cases it boils down the manners and this guy seems to at least try to be polite in his videos.
iandanforth··on Auto industry executives admit electric vehicle plans are in jeopardy
I love my plug-in hybrid. (Honda Clarity) 90% of the time I use it on electric and I never have range anxiety. When we wanted to do a cross country road trip it worked and now that we are settled I don't buy gas and don't feel bad about driving. I've owned both a Tesla and a Nissan Leaf and I probably wouldn't buy another pure EV unless it was as a second car.

Note: This sweet spot seems to have disappeared from the market as I don't believe any plug-in hybrids available in the US today match the EV range of the Clarity.

iandanforth··on Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context
https://twitter.com/DataChaz/status/1719660354743976342
iandanforth··on YouTube's ad blocker crackdown escalates, aggravating users
I got hit by this pretty hard and have since moved a lot of my viewing to https://freetubeapp.io/ (An AGPL desktop viewer for YouTube content)
← PreviousPage 11 of 34Next →