The big event of the week was Nvidia’s annual GTC conference that didn’t go by without raising a fair amount of controversy with the gaming community by using generative AI to make video games more photorealistic.
The big release of the week was, ironically, the mini and nano version of GPT-5.4 from OpenAI which is the first time we’ve seen smaller models from then since the original launch of GPT-5 last summer.
The big video of the week was Bernie speaking with Claude about AI and user privacy. It’s worth a watch as I think Claude comes across really well, but unsurprisingly it is telling Bernie all the things he’s likely to agree with!
The big Web 4.0 news is that Google is now using AI to replace headlines in search. It is not ok to change a publications headlines just to suite your needs, especially when doing so surreptitiously and without any notice or permission.
Nvidia’s DLSS 5 uses generative AI to boost photorealism in video games
This week, Nvidia had its annual GTC conference with the usual gang-buster 2 hour+ keynote from Jensen Huang. It covered everything from AI to gaming, to robotics, to Olaf from Frozen 🥶.
The most controversial announcement was gaming related where Nvidia introduced DLSS 5 which uses generative AI to make characters much more photorealistic. Needless to say that this didn’t go down well with gamers or developers and resulted in Jensen Huang just telling everyone that “they’re wrong”.
Many gamers view games as a form of art and by manipulating the graphics you’re changing the style and intent of the game designer. Yes, Nvidia say that developers and gamers are in control at all times, but to me this is another classic example of a generative AI solution in search of a problem.
Just because you can do something, doesn’t mean you should and given that many major game developers were blindsided by this announcement I see no evidence of Nvidia developing this feature in consultation with any developers or gamers.
I had started to hope that we were though the initial phase of generative AI product development where it’s just bolted onto lots (and lots!) of existing products and starting to see products that have been developed with generative AI from the ground-up. This is not one of those unfortunately.
Nvidia | YouTube | TechCrunch | Wired
OpenAI release GPT-5.4 mini and nano
We haven’t seen mini and nano model varieties from OpenAI since the initial launch of GPT-5 last summer and these update those, bringing more intelligence and the knowledge cut off up from May 2024 to August 2025.
We don’t really talk about the knowledge cutoff of models much anymore now they’re so good for searching and pulling back up-to-date information to inform their answers, but I think its still important, especially for developers who predominantly use these models ‘out-of-the-box’ and don’t necessarily give them access to live search.
Having said that, these new models are 3-4x more expensive to use that their GPT-5 counterparts, which is quite the price hike for developers. The models are definitely better, but whether they’re worth that price increase remains to be seen.
Bernie speaks to Claude
Firstly, hats off to Bernie for leaning into AI, being open-minded but critical, and paying attention to what’s happening. He has a bit to learn on how AI works as there is sycophancy at play here and Claude is telling him what he wants to hear. But, he is making a much broader, more important point about user data and privacy.
Being European, these questions seem woefully out of date. We’ve had GDPR for 8 years now specifically to try and address all of these concerns. Whilst I wouldn’t point to GDPR as a great success story in its implementation at least the fundamental idea was right - to give consumers more transparency and control over their data. However, it’s also painfully clear that GDPR is not fit-for-purpose with AI platforms and needs a re-think. I’d say that GDPR is broadly being ignored when it comes to AI platforms and that’s not a good state of affairs for anyone.
I also think this is actually quite a good ad for Claude - it comes across really well and even makes Bernie laugh with it’s answer to why consumer information is being collected - it’s simple answer is ‘Money’.
Web 4.0
Gemini task automation is slow, clunky, and super impressive
OpenAI courts private equity to join enterprise AI venture, sources say
AI Ethics News
Encyclopedia Britannica is suing OpenAI for allegedly ‘memorizing’ its content with ChatGPT
Tory chief whip reposts AI video created by far-right figure who was jailed for hate crimes
Nvidia CEO Wants Tech Execs to Stop Laying Off Workers and Scaring People
Nvidia’s version of OpenClaw could solve its biggest problem: security
Google scraps AI search feature that crowdsourced amateur medical advice
Trump’s Embrace of AI Is Blowing Up the Whole Reason He Wanted Tariffs in the First Place
Long Reads
Stratechery - Agents Over Bubbles
Shubham Bose - The 49MB Web Page
“The future is already here, it’s just not evenly distributed.“
William Gibson