KevinOnEarth @Quantillion@mstdn.io @FantasticalEconomics Dunno who #JessPhoenix is but it's succinct. 4:26 AM • December 7, 2024 (UTC)

niconiconi @niconiconi@mk.absturztau.be Q: How far are we from fusion power? A: 1 AU. 2:34 PM • December 6, 2024 (UTC)

Kyle Orland writing for Ars Technica: Longtime Valve watchers likely remember Steam Machines, the company’s aborted, pre-Steam Deck attempt at crafting a line of third-party gaming PC hardware based around an early verison of its Linux-based SteamOS. Now, there are strong signs that Valve is on the verge of launching a similar third-party hardware branding effort under the “Powered by SteamOS” label.

Isaac Callan and Colin D’Mello reporting for Global News: Auditor General Shelley Spence found as part of her annual report that, in the fiscal year ending at the end of March, the province spent a total of $104.6 million on advertising. That is the most any Ontario government has ever spent on advertising in a single year, the auditor general said, and triple what was spent the year before.

The Onion, as usual, nails it.

Ryan Christoffel writing for 9to5Mac: Live Activities are having a moment. Yesterday Apple rolled out a popular Live Activity for tracking election results via Apple News. Now, a new Safari Live Activity has been discovered in the latest iOS 18.2 beta for tracking a file’s download progress.

Kevin Purdy writing for Ars Technica: Pixelmator Pro has been advancing its capabilities recently with regular updates, including a number of AI and ML tools for adjusting photos and creating masks, vector tools, and support for more RAW photo formats and other design tool files. This is an interesting acquisition. Will Apple roll Pixelmator into Photos, or start some sort of new pro line of image/photo editing software?

Laffy @GottaLaff@mastodon.social #VOTE 1:51 PM • November 1, 2024 (UTC)

This story over at Peta Pixel compares Apple’s vs Adobe’s photo cleanup tools. Outside of some flashy animations, Clean Up works pretty much identically to Adobe Generative Remove in practice, except for the fact it will offer suggestions sometimes on objects in photos it detects and thinks you might want to remove. Otherwise, it uses the same painting method that Adobe Lightroom and Photoshop users have come to know. Since Adobe just updated Photoshop and its Firefly AI model, we figured now was a great time to see how these two widely available removal tools fare against each other. So, we tasked both with removing the same elements of six different photos to see which performed best.

May Keable 🏳️‍🌈🏳️‍⚧️ @Keab42@kind.social I think that fact that we're using AI to write emails because we find it hard and AI to summarise emails because we can't be bothered to read them suggests that we should take a look at how we communicate rather than boiling the oceans to have LLMs hallucinate at each other on our behalf. 3:12 PM • October 27, 2024 (UTC)

Picked our pumpkins today 🎃

𝐿𝒶𝓃𝒶 "not nonbiological" @Lana@beige.party EVERY STAR TREK: THE NEXT GENERATION PLOT, SORTED BY WHICH CHARACTER IS THE FOCUS: Jean-Luc Picard believes he can solve something diplomatically, immediately resorts to violence anyway William Riker believes he can solve something without violating the Prime Directive, immediately violates the Prime Directive Deanna Troi believes her Betazoid heritage won't embarrass her at work, is immediately embarrassed at work by her Betazoid heritage

Joseph Cox writing for 404 Media (Apple News): The data ultimately powering tools like Babel Street’s Locate X can come from two main sources. The first are ordinary apps installed on peoples’ phones, whose developers sell their users’ location data to a broker, who then in turn sells it either directly or through a series of middlemen to a company like Babel Street. The other is through a process called real-time bidding, in which members of the online ad industry try to outbid one another to have their advert be delivered to a certain demographic of users. A side effect is that some companies listen in on that process, and harvest location data on unsuspecting swaths of the public. 

thelooter @thelooter@chaos.social All roads lead to trains or something like that 9:47 PM • October 18, 2024 (UTC)

Reporting from New Scientist: In experiments FlipAttack was successful in extracting dangerous output 98.85 per cent of the time from GPT-4 Turbo and 89.42 per cent from GPT-4. In tests with 8 different LLMs it achieved an average success rate of 81.80 per cent.

From The National Observer: A proposal to stop labelling carbon dioxide as a pollutant and instead celebrate it as a “foundational nutrient for all life on Earth” will be up for debate at the United Conservative Party’s annual general meeting in November. I just, I don’t understand. Wasn’t Alberta on fire just a few months ago? Do they want more of that? This is just willful ignorance at this point and anyone who doesn’t understand the science of climate change should not be in any positions of power.

Adrianna Tan @skinnylatte@hachyderm.io Found this in a discussion about Strava but applicable to all tech companies #AI 5:51 AM • October 18, 2024 (UTC)

Ryan Christoffel writing for 9to5 Mac: For example, there are four requirements the FTC is imposing on all subscriptions. They must: Clearly and accurately disclose all material facts Clearly disclose the fact that the subscription will continue until cancelled Get “express informed consent” to automatic renewal Make it as easy to cancel the subscription as it was to sign up Hopefully something like this makes its way to Canada soon.

The tested LLMs fared much worse, though, when the Apple researchers modified the GSM-Symbolic benchmark by adding “seemingly relevant but ultimately inconsequential statements” to the questions. For this “GSM-NoOp” benchmark set (short for “no operation”), a question about how many kiwis someone picks across multiple days might be modified to include the incidental detail that “five of them [the kiwis] were a bit smaller than average.” Adding in these red herrings led to what the researchers termed “catastrophic performance drops” in accuracy compared to GSM8K, ranging from 17.5 percent to a whopping 65.7 percent, depending on the model tested. These massive drops in accuracy highlight the inherent limits in using simple “pattern matching” to “convert statements to operations without truly understanding their meaning,” the researchers write.

Joan Westenberg has a good piece about how she uses Reminders, Notes and Numbers as her productivity system. I really liked this part: I have almost all my information set up in a single Tasks list, split out into Smart Lists by tags. Which means adding in any information is simple as hitting Command-N anywhere in Reminders, and including a #tag in the title of the note. That, plus the natural language processing, makes capturing information a breeze.