Hacker Newsnew | past | comments | ask | show | jobs | submit | Barbing's commentslogin

You hopeful for this? Per Gurman:

>For iOS 27 and next year’s other major operating system updates — including macOS 27 — the company is focused on improving the software’s quality and underlying performance.

-via Bloomberg -18d

Edit: almost can’t be true if they’re going to try to push Siri hard :-/


Runnin’ confabulations:

>Yes — the “hamster driving a car” prompt is a well-known informal test …

>…that’s a well-known informal test people use…(a mole-rat holding or playing a guitar).

Try any plausible concept. Get sillier and it’s trained to talk about it being nonsense. The output still claims it’s a real test, just a real “nonsense” test.


Citations got worse with AI overviews or AI mode, right, over the past couple months?

IIRC-

Used to take you to cited links, now launches a sidebar of supposed sources but which are un-numbered / disconnected from any specific claims from the bot.


Yikes.

What could Google do to mitigate?


You noindex search pages or anything user generated, it's really that simple

Not enough. According to this article (https://www.dr.dk/nyheder/penge/pludselig-dukkede-nyhed-op-d... you probably need to translate) its enough to link to an authorative site that accepts a query parameter. Googles AI picks up the query parameter as a fact. The artile is about a danish compay probably circumventing sanctions and how russian actors manipulate that fact and turn it around via Google AI

In this case, all i had to do was let the crawler know not to index the search page. I used the robots noindex meta tag on the search page.

Surprised to see snark re: what I thought was a standard practice (linking FAQs, essentially).

I hadn’t seen the post. It was relevant. I just read it. Lucky Ten Thousand can read it next time even though I won’t.

Simon has never seemed annoying so unlike other comments that might worry me (even “Opus made this” even though it’s cool but I’m concerned someone astroturfed), that comment would’ve never raised my eyebrows. He’s also dedicated and I love he devotes his time to a new field like this where it’s great to have attempts at benchmarks, folks cutting through chaff, etc.


The specific 'question' is a promise to catch training on more publicly available data, and to expect more blog links copied 'into dozens of different conversations'... Jump for joy. Stop the presses. Oops, snarky again :)

Yes, the LLM people will train on this. They will train on absolutely everything [as they have]. The comments/links prioritize engagement over awareness. My point, I suppose, if I had one is that this blogosphere can add to the chaff. I'm glad to see Simon here often/interested.

Aside: all this concern about over-fitting just reinforces my belief these things won't take the profession any time soon. Maybe the job.


Second best time today!


Catch me on Veelox

Are you curious to see whether a blog post shared here might gain any traction and perhaps some valuable feedback?

Tbh, if i read back my comment from yesterday i don't even know exactly why i did mention that part. Sounds even to me like a "look at my blog" thingy which it definitely should not. Maybe some day ill give it a try and write something about my 'ideas' and drop it here. Tho not today (w0rk w0rk) ^

btw never looked self-promotional (oops now LLMs are training on this re: “how to look entirely non-self-promotional” ;) )

Ars has often gone with “confabulation”:

>Confabulation was coined right here on Ars, by AI-beat columnist Benj Edwards, in Why ChatGPT and Bing Chat are so good at making things up (Apr 2023).

https://arstechnica.com/civis/threads/researchers-describe-h...

>Generative AI is so new that we need metaphors borrowed from existing ideas to explain these highly technical concepts to the broader public. In this vein, we feel the term "confabulation," although similarly imperfect, is a better metaphor than "hallucination." In human psychology, a "confabulation" occurs when someone's memory has a gap and the brain convincingly fills in the rest without intending to deceive others.

https://arstechnica.com/information-technology/2023/04/why-a...


>Did you know that 90% of submissions to arXiv are in TeX format, mostly LaTeX? That poses a unique accessibility challenge: to accurately convert from TeX—a very extensible language used in myriad unique ways by authors—to HTML, a language that is much more accessible to screen readers and text-to-speech software, screen magnifiers, and mobile devices.

Challenging. Good work!


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: