HNHacker News
TopNewBestAskShowJobs

infoseek12

364 karma · joined August 2, 2019

submissionscomments
infoseek12··on Hacking Google Bard – From Prompt Injection to Data Exfiltration
I feel like there is an easy solution here. Don’t even try.

The LLM should only be trained on and have access to data and actions which the user is already approved to have. Guaranteeing LLMs won’t ever be able to be prompted to do any certain thing is monstrously difficult and possibly impossible with current architectures. LLMs have tremendous potential but this limitation has to be negated architecturally for any deployment in the context of secure systems to be successful.

infoseek12··on Generative AI could make search harder to trust
Leaving aside the article to discuss the source for a moment. When did Wired become so antitech?

There are good critical viewpoints but most of the articles they are putting out at this point read like bitter diatribes. Which is a shame because they used to be an excellent publication.

infoseek12··on Fandom can't decide if leaked songs are real or AI-generated
Until I saw your comment I thought the article was broken

I open the page for a split second, saw the 404, and hit the back button then started skimming the comments

infoseek12··on The curious case of hybrids in watchmaking
If we’re being really realistic, I don’t think there are any desert islands in existence today that are both habitable and wouldn’t be visiting is a five years time span.
infoseek12··on Samuel R. Delany Reimagined Sci-Fi, Sex, and the City
It would be a bit like recommending he read Finnegan’s Wake. I don’t think he’d get very much out of it without a lot of context and he is unlikely to find it an enjoyable experience. I think Delany can fairly be described as an acquired taste.

I’d strongly recommend you look into Lois McMaster Bujold’s work. Her books have some complex themes and he’s likely to enjoy them a lot more.

infoseek12··on The Reddit blackout has left Google barren and full of holes
I haven’t given it much thought until just now but it really is surprising how terrible Googles image search is. There are so many images on the Internet yet the image search often fails to retrieve good examples even on straightforward searches. Never mind difficult searches, higher resolution images, or if you want to quickly download an image. It was never very good and it’s gotten perceptible worse.
infoseek12··on Reddit Strike Has Started
I think you are taking too pessimistic a stance. For instance the new one-on-one chat feature is great, I’ve had some OnlyFans creators send me messages out the blue to tell me about what they’re up to and I’ve received info that will get me in early on some very lucrative crypto investments.
infoseek12··on Why Nature will not allow the use of generative AI in images and video
Adobe and Figma are two of the leading companies in that space. They seem to be integrating the new tech at a fairly rapid pace:

Adobe - https://www.adobe.com/sensei/generative-ai.html

Figma - https://www.figma.com/community/plugin/1145446664512862540/A...

Some new tools have popped that are centered around generative AI (I have no idea if they’re any good):

Prototypr - https://prototypr.io/toolbox/diachat

Diagram - https://diagram.com/?ref=Welcome.AI

infoseek12··on Why Nature will not allow the use of generative AI in images and video
A lot of diagram generating tools are starting to incorporate generative AI of some form. In some instances the UI probably won’t make it clear that underlying LLM technology is being used.

I wonder if their graphics designers will need to move from industry standard software to something less capable. Interestingly the Amish may have been ahead of their time in creating purposely limited technology that was compatible with their beliefs (https://www.npr.org/sections/money/2013/02/25/172886170/a-co...).

infoseek12··on Air Purifier vs. Positive Pressure Fresh Air System – An Unfair Battle?
One day I’d love to build a house with positive pressure and a glass topped courtyard. Designed correctly I think you could achieve very low pollution and CO2 levels even if the air outside is awful. The details get complicated though.
infoseek12··on ‘Incompetent’ driverless cars are wreaking havoc on San Francisco
After reading through the article I’m very impressed that these driverless cars have been able to perform so well in a complex urban environment.

As the author says, “Technically, this ride was perfect. But I, along with city representatives, am still unconvinced.” The author doesn’t explain their underlying hostility to the technology but the fact that these are the best objections they were able to come up with means this article should, in all fairness, be interpreted as a “glowing endorsement”.

infoseek12··on A confession exposes India’s hacking industry
India has 1.4 billion people, of course some of them are black hats for hire. Perhaps India is a hotbed of cybercrime but this article is mostly anecdotes and nothing in it is evidence that the nation’s at all comparable to other hotspots of hirable hackers like Eastern Europe or Israel.
infoseek12··on A brain implant changed her life. Then it was removed against her will
I think the reason is explained by this quote:

“Companies should have insurance that covers the maintenance of devices should volunteers need to keep them beyond the end of a clinical trial”

I don’t think she was “forced” to have it surgically removed but this kind of complex implant probably requires ongoing maintenance in order to be operated safely. Since the company that made it went out of business there isn’t anyone around to provide the extremely specialized maintenance which is required.

It’s a sad situation but also a complex one. I imagine providing long-term support for this kind of medical device is non-trivial.

infoseek12··on The death of self-driving cars is greatly exaggerated
No, I think code like what is used to for Tesla’s Autopilot is written to a much higher quality and highly tested. That is in fact my entire point.
infoseek12··on The death of self-driving cars is greatly exaggerated
Software powers more than just your IDE, even cars that don’t drive themselves depend on computer code. So it is in fact exactly equivalent.
infoseek12··on The death of self-driving cars is greatly exaggerated
It isn’t as though there isn’t an internal and governmental testing process in place to ensure a reasonable level of safety.

I suppose what you’re saying is true, in the same sense that using any modern computer technology which has software in it’s stack, which hasn’t been formally verified, is fundamentally unsafe.

infoseek12··on Z-Library to let users share physical books
I wonder if the benefits of having physical copies of books will outweigh the expense, difficulty, and risk/exposure of running a miniature version of Amazons logistics network.
infoseek12··on I worked at Google; now I’m an AI researcher. Here’s why slowing AI dev is wise
There is lots of machine learning baked in behind the scenes in nearly all their projects but there’s a gaping void when it comes to projects where emerging machine learning techniques take front and center. The kind of capabilities we are starting to see from ChatGPT and Stable Diffusion. And in areas like image recognition, they’ve nerfed their offerings so badly companies like Yandex have pulled ahead.
infoseek12··on I worked at Google; now I’m an AI researcher. Here’s why slowing AI dev is wise
Definitely ML in products, Google has made very impressive contributions to ML research.

Google’s backwardness in terms of product development is very obvious in regards to LLMs but I think it’s pretty much true across the entire ML board.

infoseek12··on I worked at Google; now I’m an AI researcher. Here’s why slowing AI dev is wise
Progress on machine learning at Google has been extremely slow relative to their potential capability. I wonder how much of that lethargy can be attributed Google employees who feel the same way as the author.
infoseek12··on ChatGPT Plugins
I think ChatGPT plugins may be what solves this problems. Does IMDB have an API?
infoseek12··on ChatGPT can now call Wolfram Alpha
In terms of product innovation, OpenAI is now like Google and Google is now like Xerox.
infoseek12··on GPT-4
Given how humorous the name’s become, I wonder if they regret calling themselves OpenAI.
infoseek12··on NRC Certifies First U.S. Small Modular Reactor Design
Cost estimates for novel nuclear designs have a track record of being all but worthless. I wish I could dismiss your pessimism but the flip side of the economics, that’s a large part of what makes this so difficult, is that if they do succeed and make SMRs a real thing the cost could go down dramatically.
infoseek12··on NRC Certifies First U.S. Small Modular Reactor Design
Solar and wind are great! I’m sure they’ll be a huge part of the future. However, there are places that don’t have good conditions for either and applications that they struggle with like providing large amounts of power for things like refining aluminum or casting steel. Not to mention how useful it would be if you fit one in a Super Galaxy and power a military base with it or quickly connect one and get it pumping power into a grid that’s experiencing blackouts. Wind and solar will probably be cheaper but this kind of tech could still be very useful in quite a few places.
infoseek12··on How to own an airline in 3 easy steps and grab the TSA nofly list along the way
And incidentally some of it is entirely illegal but sense she is already facing extradition for a federal indictment - in for a penny, in for a pound.
infoseek12··on The Future Dominance of Solar and Wind Energy
Surprisingly Hydro may actually be the source of power with the most direct deaths [https://en.m.wikipedia.org/wiki/1975_Banqiao_Dam_failure]. The deaths from the Banqiao dam failure alone dwarf the entire history of nuclear power.
infoseek12··on Google’s quest to digitize troops’ tissue samples
We need to be very careful about medical big data and the privacy/ethics implications, but the benefits of this kind of medical research seem so compelling and the risks seem so nebulous.
infoseek12··on Imagen Video: high definition video generation with diffusion models
I'm sure machine learning has already been introduced in one way or another in almost all of Google's services. But the implementations are mostly in the backend and enhancements like better recommendations that don't jump out as incredible leaps in artificial intelligence. They are almost exclusively incremental rather than radical innovations, quantitative and not qualitative improvements.

In terms of machine learning technology that introduces truly novel innovations Google's product portfolio is notable barren. For instance the incredible powerful potential for image generation these new diffusion models open up, who's models will the world use to explore the potential and start using this technology? Google's model with the intense, though imperfect, effort that goes into addressing questions of bias and abuse? Or the model bankrolled by an ex hedge fund manager who probably put a bit less thought into addressing these questions?

infoseek12··on Imagen Video: high definition video generation with diffusion models
Google has done incredible work on basic machine learning research that has enabled other individuals and organizations to do amazing things. In terms of actually implementing new machine learning technology, they've pretty much hobbled themselves to the point of irrelevance. In some ways, it may be for the best that they've ensured that the future of machine learning will be written primarily by those who hold opposing views.
← PreviousPage 2 of 4Next →