Working at this scale and price tag, it seems like it would be easier to install a small robot arm in the floor to do it. Or a setup like a 3-axis milling machine, just to align with a port on the bottom and plug it in.
I guess that doesn't have the sexy Nikola Tesla factor.
I recently watched Not Just Bikes' video on the disastrous future side effects of self-driving cars[0]. Of course it made me think about the massive PR push that made us think they were around the corner, but also about the manufactured consent for these technologies in the first place. Right now this kind of discussion is hitting the mainstream with the 'clanker'[1] backlash. I think it's really obvious to a lot of people that the AI push is not organic and is not based around consumer needs, and this manipulation is making people genuinely angry[2] (ok jreg is a performance artist, but just because something is performative doesn't mean it's not real).
After I set up a self hosted git forge a little while ago, I found that within minutes it immediately got hammered by OpenAI, Anthropic, etc. They were extremely aggressive, grabbing every individual file from every individual commit, one at a time.
I hadn't backlinked the site anywhere and was just testing, so I hadn't thought to put up a robots.txt. They must have found me through my cert registration.
After I put up my robots.txt (with explicit UA blocks instead of wildcards, I heard some ignore them), I found after a day or so the scraping stopped completely. The only ones I get now are vulnerability scanners, or random spiders taking just the homepage.
I know my site is of no consequence, but for those claiming OpenAI et al ignore robots.txt I would really like to see some evidence. They are evil and disrespectful and I'm gutted they stole my code for profit, but I'm still sceptical of these claims.
Cloudflare have done lots of work here and have never mentioned crawlers ignoring robots.txt:
I was thinking about this for the recent GMTK game jam, which crashed itch.io.
Is it really such a bad thing for your launch/event to crash a platform? Nobody is going to decide not to buy Silksong after all because it's so wildly popular it brought down Steam. It generates a great deal of positive headlines. To me it seems like a good problem to have.
This study itself and also the media coverage of it are shockingly bad. I wrote a bunch about it at the time and I don't really want to do that again but here is the low down:
- This is not a longitudinal study. Each partipant did 4 20 minute sessions. It just happens that the total study took 4 months.
- The paper does not imply long term harm of any kind, they just measured brain connectivity during the short tasks.
- It is not surprising that when asked to use an LLM to write an essay, partipants don't remember it. They didn't write it.
- It is not surprising they showed less brain activity. They were delegating the task to something else. They were asked to.
- I think the authors of the paper deliberately attempted to obscure this. Q7 on p30 is "LLM group: If you copied from ChatGPT, was it copy/pasted, or did you edit it
afterwards?" This has been removed from the results section entirely, and other parts of the results do not match the supposed methodology.
- The whole paper is extremely sloppy, with grammar mistakes, inconsistencies, and nonsensical charts. Check out Figure 29...
It's just horrible. For me the preview doesn't show until I've hovered for ~3 seconds (even before I turned on AI), in which time I could have long since middle clicked the page, skimmed, and closed it again.
This is the content for the preview:
> www.mozilla.org
> What's new with Firefox 142
> What's New | Firefox 142
> 3-4 mins reading time
No OpenGraph descriptions are good, so I can't see it ever being better than this. I don't know why this reading time metric has become a thing, it's useless because it doesn't know which parts of the page I'm interested in. I could actually see the full url from the immediate link preview, so having only the domain here is worse than useless.
The AI summary is both too short and too slow to be useful (unless maybe you're running an RTX 6000 or whatever). For this link, it only mentions Relay.
And even the basic behaviour seems broken. The preview appears at seemingly random locations on the page, sometimes under the cursor and sometimes far below. When it does decide to appear away from the cursor, releasing the mouse button actually follows the link, completely negating the purpose of the preview!
In this case it's kind of unbelievable how quickly the flip happened. Their 'manifesto'[0] was published only at the start of the year, where they vowed to be an incorruptible bastion of stability in a selfish society focused on short-term thinking. They want Bear to be around in 50 years.
Then in the last couple weeks or so[1] it seems they saw a bit of a spike, and immediately pulled up the ladder.
They even criticised this sort of behaviour in their manifesto:
>We've seen our fair share of open-source projects become sour (see the recent Wordpress drama) or abandoned entirely. We've seen OpenAI become ClosedAI. There's a common thread here. Trust isn't just a legal structure, but a social contract.
I am actually totally in favor of source available licenses, but in this case it seems counter to all the boasts the developer has made about their platform.[2]
reply