Beep boop - this is a robot. A new show has been posted to TWiT…
What are your thoughts about today’s show? We’d love to hear from you!
Great show, long time fan. Keep up the good work. Btw do you publish the show notes “the rundown” if so, where can I view it? I know you don’t have time to cover every topic of the week but I would still like to see the bits that don’t make it on to the show. Thank you!
1 Like
Regarding the AI analyst, if the AI has been certified and it is compliant with the data protection laws, and, for example HIPPA in the USA, there should be no problems with using the AI for psychiatry, but if it isn’t certified and it isn’t compliant with data protection and medical rules, it shouldn’t be used.
Regarding Perplexity ignoring robots.txt, I am with Paris for the most part and, sorry Jeff and Leo, Perplexity going to the site and reading it for you and giving the summary is not the same as you going to the site with Firefox, Chrome etc. that argument is so obviously specious, I was very surprised that either of them said it, let alone both. In the case of a web browser, the site is displayed as the site owner/creator intended, if Perplexity goes to the site and scrapes the information for you, it isn’t.
Also, “everything” should be readable, again, I am with Paris, not for a corporation making money out of it, that company needs to pay for the rights to use that information, unless it is explicitly given away, such as under creative commons or public domain.
There are often parts of a site that should not be scraped, so ignoring robots.txt is not just impolite, it is going against the wishes of the sites owner. For example, the backend CMS pages should never be scraped, likewise, the images should be scraped in context, not scraped as a directory, if the owner so wishes.
What we really need is an AI broker, that scrapes websites regularly, signs the licensing agreements with the site owners, pays them, then resells the collected database of information to the AIs. This would remove their need to continually scrape sites, killing their performance and causing unreasonable costs to the site owners due to the excessive traffic they cause - the AI scrapers should paying the site owners for that reason alone; the reports say that the AIs are much more aggressive that normal search spiders, coming to the site much more often than is really necessary.
2 Likes
As a web developer I agree with this wholeheartedly. Leo and Jeff are assuming that the AI is merely serving you the content. Does anyone believe that to be the case?
There’s decades of precedent for companies blocking a certain web browser, and the reasoning boils down to a business one. We could argue that AI’s are just another web browser and if my company wants to block those, they’re free to do so. Is that a good idea from a business standpoint? Time will tell.
Right to read applies to end users and only end users. This notion that we have to let AI’s gobble up everything and nobody can say otherwise is utter horse 
.
1 Like
Also I respect that different people grieve in different ways but I don’t see anything healthy about talking to a fake hologram of your loved ones.
1 Like
Regarding the AI-generated obituaries, I could see that being a help for families who don’t have confidence in English because it’s not their primary language, but only if the AI can be guaranteed to never hallucinate. And native-speaker human review would still be necessary before publication.
As for AI describing a website, it’s important to distinguish between AI summarising content ‘in its own words”, which alters the content, and assistive technologies for people with sight limitations, which read out the content as written, but may have to be somewhat descriptive to help the user understand how the content is presented to a sighted user. It could be a bit of a grey area, especially once you start talking about restricting what technologies can be used.
I could also see an an AI summary potentially being useful to help someone with cognitive limitations understand a complex site, and decide which part of the site is relevant to their needs.
1 Like
Thanks for asking. I don’t publish the show notes, exactly, but the raw materials, at least from me, are posted in two places.
On Mastodon, TWiT News Feed (@twitnews@twit.social) - TWiT.social That feed updates in real time as I bookmark things. The tag at the end of the toot shows which show’s rundowns I intend to add the story to.
My Raindrop.io account is public, and all the links go there, too:
Again, there will be tags on every story which indicates where I want them to go.
Our workflow is powered by the Raindrop page. Zapier copies each entry into a Google Sheet called Leo’s Links. The show producers then copy those links into the official rundowns. We don’t publish those because they contain a lot of extraneous material.
Some of the shows have their own show notes. Steve keeps the Security Now notes on his website. The Windows weekly show notes are in Notion but Kevin usually posts them into the Club TWiT Discord during the show. MacBreak Weekly has my links, and Andy Ihnatko basically produces his own show notes which he will post on Ihnatko.com if it ever opens. I think he also does a YouTube recap after each show.
2 Likes