Hi
Hope you had wonderful weekends wherever you may be. We’re counting down to Independence Day here in Vietnam and flags are popping up everywhere – visually awesome and we’ve all got a long weekend to look forward to #winning
A few notes from the world of AI to start your Monday:
AI Scaling Forecast: Trillion-FLOP Titans by 2030?
Hold onto your GPUs! A new study from Epoch AI suggests that we’re just getting started with AIs and they could get a lot smarter yet. The study suggests AI training runs could hit a mind-boggling 2e29 FLOPS (what on Earth is that?) by 2030 – 10,000 times more powerful than today’s top models. To put this in perspective, an AI system this powerful is loosely equivalent to having the combined brain power of billions of people, working non-stop at superhuman speeds. Wild stuff…. that said, researchers identified four key bottlenecks: power supply (those data centres are thirsty), chip manufacturing (silicon shortages, anyone?), data scarcity (even the internet has limits), and the ‘latency wall’ (#physics). Despite these hurdles, the AI express train shows no signs of slowing down. 🤖💡
Either way, the tech titans are already placing their bets. Mark Zuckerberg’s Meta is splashing out billions on chips for AI in 2024 alone, while Sam Altman of OpenAI is reportedly exploring investments up to a jaw-dropping $7 trillion in the quest for AGI. With rumours of $100 billion data centre projects swirling, it’s clear some players are all-in. But here’s the real question – even if we can build these AI behemoths, should we? As we race towards 2030, balancing innovation with ethical considerations will be crucial. One thing’s for sure: the AI race is heating up, and the finish line just got a whole lot further away. 🏁💰
OpenAI’s Condé Nast Deal: AI News Discovery Gets Glossy
OpenAI’s just inked a deal with publishing powerhouse Condé Nast, bringing a touch of glamour to AI-driven news discovery. The partnership will see content from iconic brands like Vogue, The New Yorker, and Wired popping up in ChatGPT and the intriguing SearchGPT prototype. This latest move adds Condé Nast to OpenAI’s growing list of media partners, including the Associated Press, News Corp, and TIME.
It’s also a clever play to bolster SearchGPT’s credibility and potentially sidestep those pesky copyright lawsuits (and a real twist after their admission earlier this year to the UK House of Lords its “impossible” to train AI without copyrighted materials 🤔). But with OpenAI’s admission about copyrighted content, the legal battles are far from over. Will these partnerships be enough to satisfy the courts, or is this just the opening act in a long-running AI copyright drama? Stay tuned – this story’s got more twists than a New Yorker short story. 🍿👀
AI Reading List: A Roadmap for the Curious and the Concerned
Looking for some light AI poolside reading for the Independence Day long weekend? The UK’s House of Commons Library has released an extensive reading list on AI, offering a valuable resource for policymakers, researchers, and the public. A pretty comprehensive guide, the reading list ranges from basic definitions to discussions of complex ethical issues.
Framed by the UK’s current “context-based” regulatory approach, the list dives into AI’s impact across various sectors, including healthcare, finance, and creative industries while also diving into critical issues such as ethics, safety, and potential risks. And with the new Labour government signalling a possible shift towards more direct regulation of powerful AI models, this reading list provides some very timely insights into the evolving landscape of AI governance. 🧠💡 #AIPolicy #TechLiteracy
CRAFT-ing the Future: Danny Liu’s Blueprint for AI in Education
An interesting framework for tackling generative AI in HE from Danny Liu at the University of Sydney. Dubbed CRAFT: Culture, Rules, Access, Familiarity, and Trust, it’s an attempt to provide a structured approach for educators grappling with the AI wave – and it’s not just about slapping AI tools into curricula and calling it a day; it’s a more nuanced approach:
CRAFT starts with considering the Culture needed for responsible AI experimentation, then moves to establishing clear Rules (goodbye, confusing “AI-free” pledges). Access is a key focus – Liu advocates for widespread availability of advanced AI, not just for the tech-savvy. Familiarity involves understanding AI’s capabilities and limitations, while Trust explores the dynamics between students, educators, and AI systems. Liu’s not just theorising; you’ve probably heard of Cogniti, a DIY AI platform adopted by over 70 institutions and counting (overview video here). While it’s too early to call CRAFT a silver bullet, it’s certainly a framework worth exploring for those looking to thoughtfully integrate AI into educational settings. 🛠️🎓
Mad Max Meets the Muppets: AI’s Wild Ride on Furry Road 🎭🏜️
A bit of a change in pace from the Runway Gen:48 competition we mentioned last week, this is an awesome AI mashup – Mad Max: Muppets. Using a potent blend of GPT-4 and Midjourney V5 & V6, they’ve conjured up a world where Kermit rules as Immortan Joe and Miss Piggy brings the fury as Furiosa. With prompts like “Film still of a Muppet as a Muppet War Boy driving a modded car in the desert”, this technicolour fever dream isn’t just technically impressive – it’s a masterclass in creative collision. Mad Max: Furry Road! 🐸🐷💨
We hope this edition of the newsletter has been of interest to you. If you’re new here and it’s been useful, please do click subscribe and you can expect a weekly update every Monday from now on. If you’re already a subscriber – thanks for your ongoing interest and support! Either way, if you know others who might benefit from reading this weekly, please forward it on to them too.
Have a great week ahead and let us know if there’s anything we’re missing that we should add to make this newsletter more useful for i) yourself and/or ii) others. This is a fast-moving, ever-evolving space and we greatly value any and all feedback. 🙏