There’s no doubt that we have the cart before the horse in terms of showing that these models can perform without fully being able to explain why. I think it’s the fact I agree with you on this that makes me sensitive to the areas where we actually do have explanations.
It’s the principle of doing the scientific research before making the claim. Trust me, I am frustrated by how anti-scientific much of the “research” coming from the AI industry is (Anthropic “publishes” “papers” that have not undergone any form of peer review) - let’s be better.
Happy to do it - I think it’s a really interesting question and worth a review either way. If I provide you with those attempts, will you retract your claim that nobody has an explanation?
Would you like me to do a short literature review to evaluate the merits of the claim you made?
(I’m being facetious - I don’t think you did because a cursory google search found that in fact, someone has an explanation)
I just thought your significant claim that “nobody has any explanation” might warrant a review of the literature. Did you happen to do one?
Your personal ignorance of the science actually undermines the point you are looking to make.
Real research explains the phenomenon, but also shows it harms accuracy: arxiv.org/abs/2603.18507
Apps like tinder will actually penalize you for swiping right as fast as you can on everyone. But that doesn’t mean they’re giving anyone a fair shake - they effectively decide how many matches you’ll get in a given period before you even interact. It’s all about making you pay for premium.
Yikes, ventilation must be really bad at the revue. Are those CO2 levels safe? A mask will protect you from disease but not CO2
Later in the article she says she felt safer getting her children an unlicensed vaccine from “BabyJabs” than the licensed vaccine from the NHS
New research finds that car touchscreens are so distracting that they induce "significant performance degradation" among drivers.
Meanwhile, in-vehicle touchscreens remain completely unregulated in the US.
doi.org/10.1145/3746...
If I’m reading this correctly, *directly next to Dufferin mall* is considered a “neighbourhood corner store”?
Do the existing e-bikes support it (maybe with modification)? Or would they need to get entirely new bikes
Now we just have to hope it never gets hot in May!
I asked @juliedzerowicz.bsky.social before the election whether the feds approve U.S. military participation in the Toronto Airshow. Her office just replied saying they can’t answer because Cabinet hasn’t been named, and to ask again “in a few months” (after the airshow?)
They have been doing that for a long time. Early versions of GPT used Reddit comments sections to judge the quality of a response during model training.
Although this is recommended, it’s actually not allowed for a car to run over a pedestrian because they aren’t pointing.
Cars are significantly heavier than bicycles, and bring several orders of magnitude more energy into a collision even at the same speed as a bike.
Provinces can’t unilaterally secede - see here: www.canlii.org/en/ca/scc/do...
“secession of a province "under the Constitution" could not be achieved unilaterally, that is, without principled negotiation with other participants in Confederation within the existing constitutional framework.”
The "leading" numbers really don't mean much. This is evidence of that
You’re absolutely right about how image “generation” works but it is a little different for text since it’s word by word. The quoted post is right, though, LLMs are trained to predict sequences of words that seem likely - nothing to do with truthfulness
It’s like delivery bike riders are completely invisible. Saying “it’s all delivery e-bike riders” as if that means it’s not worth it - okay, is everyone going to stop using DoorDash too?
its amazing how chatgpt knows everything about subjects I know nothing about, but is wrong like 40% of the time in things im an expert on. not going to think about this any further
Philip Mai PoliDashboard Making the Invisible Visible Civic Tech TORONTO Hacknight #488 IN PERSON AND ONLINE Tues April 8 → 7pm
Next Tuesday, see @philipmai.com talk about PoliDashboard!
Discover how political advertisers in Canada are chasing your clicks on Facebook and Instagram, and how political ads can be weaponized to manipulate, or interfere in an election.
8 April 7pm, in person/online
guild.host/events/civic...
From News to Action: How The Green Line's Documenters Are Changing Civic Engagement.
Anita Li and Sebastian Tansil discuss how local news can empower communities to take action.
Plus our usual breakouts for projects and discussions!
March 18th 7pm: in person and online
guild.host/events/civic...
Documents unearthed as part of a legal challenge show the Doug Ford government received expert evidence that planned bike lane removals could worsen traffic congestion in Toronto and lead to injuries.
#BikeTO #ToPoli #OnPoli
There is also an incredible amount of talent specifically focused on civic tech — technology that seeks to make the city work better — with a Civic Tech Toronto group meeting weekly to show off cool independent projects, often built using city hall’s generally unheralded but very useful Open Data department.
Great to be mentioned in the Toronto Star this morning - thanks @graphicmatt.com!
www.thestar.com/opinion/cont...
It is!
hey if you aren't american i'd strongly reconsider traveling to the US for any sort of convention
