AI and digital freedoms in Canada https://xitter.cf/t.co/JRCRa4b4zh

Vancouver British Columbia
Joined November 2010
Why have we heard so much about Kimi K3 and open-weight models this week? What do they mean for the future of AI? K3 caused a stir by matching the benchmarks for closed frontier models like GPT-5.6 and Opus 5 — but it’s free to download. In the Globe last week, I argued that more of our AI use will shift to smaller models we run locally, and thus not in big data centres we’re in such a rush to build out. K3 isn’t small, but it points to how fast open-weight models are catching up. In today’s post, I sketch how open-weight models are changing the landscape, pointing to a future where we run smaller models on our own devices, with more privacy and control, and less impact on the environment. robertdiab.ca/posts/open-wei…
1
99
Lead piece in the NYT: “Last year, the [data centres] consumed 64 gigawatts of electricity globally, roughly as much electricity as Germany consumes… By the end of 2030, that is expected to quadruple, eclipsing the power used by all countries in South America and Africa combined.
1
67
Open-weight models are not open source. Much about them is never disclosed. Canada’s consultation on AI transparency will look at many issues, including how much model makers should have to disclose. robertdiab.ca/posts/ai-trans…
89
“Choosing a direction shouldn’t be left to individual professors tinkering on their own, which is what is happening now. Achieving the right balance requires redesigning entire degrees, assessments, timetables and hiring – decisions only presidents, provosts and boards can make.”
101
Good catch. Wonder if others will follow.
Kimi K3's weights have been released with a more restrictive license than before. - Companies operating a Model-as-a-Service (MaaS) business with over US$20 million annual revenue must negotiate a separate commercial agreement before using the model commercially (similar to MiniMax's modified license since M2.7) - Products over 100 million MAU or $20 million monthly revenue must display "Kimi K3" prominently - Large providers are exempt from the commercial restriction if use is purely internal huggingface.co/moonshotai/Ki…
1
1
117
Canada has launched a public consultation into AI transparency, with a discussion paper that looks at: -AI-generated content; -AI tools; -Information about the models; -AI incidents; and -AI agents. Will more disclosure make AI seem more trustworthy? robertdiab.ca/posts/ai-trans…
56
Robert Diab retweeted
This is great. Stop the slop.
Today, Substack is launching an AI detection feature, via an integration with @pangram. Going forward, you’ll be able to scan posts, replies, and comments on the Substack app to see an estimate of how much of it was written by a human, or with AI assistance.
3
6
44
8,906
Many insights in Reuter's Digital News Report 2026, including its main finding: platforms (TikTok, YouTube) are — for the first time — more popular sources of news, on a global average, than TV and legacy news venues. More in today's post: robertdiab.ca/posts/who-cont…
66
Robert Diab retweeted
Senator Wyden warns that Bill C-22 represents a “counterintelligence threat” and urges the administration to take action. Given the latest Trump tariff threat, lawful access could add yet another complication in the increasingly fraught trade relationship. michaelgeist.ca/2026/07/rush…
4
28
1
63
2,699
Just my two cents: the census questions do not engage our right to privacy under the Charter. They touch on private matters, but they don’t do so for a law enforcement or intestigatory purpose. They don’t constitute a search or seizure in the same way that mandatory reporting of your income under the Income Tax Act isn’t either.
Can the government compel Canadians to answer deeply personal questions? Constitutional lawyer Hatim Kheir explains the Federal Court challenge to the 2026 long-form census, where lawyers funded by the Justice Centre argue that mandatory disclosure of sensitive personal information violates Canadians' Charter rights. youtube.com/watch?v=whBX7RQZ…
133
Gregory Conti in Compact Magazine: "the most vocal skeptics have not fully reckoned with how anthropomorphic AI—models that mimic human personality traits, emotional states, and patterns of thought, as currently commercially available generative AI models do—is currently in the process of unsettling human psychology and the social fabric." compactmag.com/article/the-a…
66
Can you just ignore AI and still be a competent lawyer? (or law prof?) Or will all lawyers soon need to use AI in some capacity to be competent? I take this up at slaw.ca, making the case that soon, most of us will need to be on board with AI. lnkd.in/gtpYyhm8
1
73
Another insight from The Atlantic piece on reading — the few of us still reading form a small strong subculture. But should we be consoled by this? theatlantic.com/magazine/202…
2
1
2
15
1,543
An exceptional contribution to the Gutenberg Elegy tradition, with many unsettling stats, including this one: “A study analyzing 236,000 responses to the American Time Use Survey found that the the proportion of Americans who read for pleasure on any given day fell from 28 percent in 2004 to 16 percent in 2023” theatlantic.com/magazine/202…
3
694