The true hazard isn’t that machines will act with out oversight; it’s that human overseers do not know what the machines are literally “pondering.” Fortunately, science could provide a approach ahead.
Learn the total op-ed on the pressing want for brand spanking new safeguards round AI warfare.
The must-reads
I’ve combed the web to seek out you in the present day’s most enjoyable/essential/scary/fascinating tales about expertise.
1 Regardless of blacklisting Anthropic, the White Home desires its new mannequin
Trump officers are negotiating entry to Mythos. (Axios)
+ Anthropic stated it was too harmful for a public launch. (Bloomberg $)
+ Finance ministers are alarmed concerning the safety dangers. (BBC)
+ Anthropic simply rolled out a mannequin that’s much less dangerous than Mythos. (CNBC)
+ The Pentagon has pursued a tradition battle in opposition to the corporate. (MIT Expertise Overview)
2 Sam Altman’s facet hustles have raised conflict-of-interest issues
His opaque investments might affect selections at OpenAI. (WSJ $)
+ A jury will quickly determine if OpenAI deserted its founding mission. (Wired $)
+ The corporate is making an enormous play for science. (MIT Expertise Overview)
3 A Starlink outage throughout drone checks uncovered the Pentagon’s SpaceX reliance
It was certainly one of a number of Navy take a look at disruptions linked to Starlink. (Reuters $)
+ The DoD can also be tapping Ford and GM for army improvements.(NYT $)
4 Information heart delays threaten to choke AI enlargement
40% of this yr’s initiatives are susceptible to falling delayed. (FT $)
+ Partly as a result of nobody desires an information heart of their yard. (MIT Expertise Overview)
5 Alibaba simply launched its personal model of a world mannequin
Comfortable Oyster is the newest try to increase AI’s means to understand bodily actuality. (SCMP)
+ However they nonetheless want to know trigger and impact. (FT $)
6 Google’s Gemini is now producing AI photos tailor-made to private information
By analyzing customers’ Google companies and information. (Quartz)
+ Google says it would minimize the necessity for detailed prompts. (TechCrunch)
7 OpenAI is beefing up its agentic coding and improvement system
Its Codex replace is a direct shot at Claude Code. (The Verge)
+ However not everyone seems to be satisfied about AI coding. (MIT Expertise Overview)
8 Europe’s on-line age verification app is right here
It’s out there free of charge to any firm that wishes it. (Wired $)
9 Smartglasses are giving Korean theaters hope of a Okay-Pop second
Their AI-powered translations are taking the exhibits to the world. (NYT $)
10 World voice actors are preventing Hollywood’s AI push
Their voices are coaching the fashions which might be changing them. (Remainder of World)
Quote of the day
“There’s this darkish interval between now and a while sooner or later the place the benefit may be very a lot offensive AI.”
—Rob Joyce, former director of cybersecurity on the Nationwide Safety Company, tells Bloomberg how AI is creating new hacking threats.
One Extra Factor