SAIL: OpenAI needs adult supervision, Nvidia nails it, Tutors, LearnLM, Jobs
Sensemaking, AI, and Learning (SAIL)
AI is facing increased reckoning from regulatory interests, with EU leading. In the land of significant freedom, OpenAI rules the LLM space, as last week’s GPT-4o cemented. Worth noting, most people I know have seen the video. But after a week, it has 4m views. The AI conversation bubbles into public view sometimes, but largely, it’s happening in a small segment of society. The real issue these days is OpenAI’s rather consistent habit of owning themselves. They did it with the firing (and rehiring) of Altman. They did it with employee investment plans. Then with eliminating their “super alignment team” (safety). And now with the voice powering their omni model. Given how potentially significant their impact on society will be, they need adult supervision.
AI and Learning
Sal Kahn hosted a discussion with co-founder of OpenAI, Greg Brockman on mostly education. At the 14min mark, the focus turns to tutors, the current low-hanging fruit in AIed.
More here on a Khan Academy announcement with Microsoft. The entrants into AI in education is still open, but Pareto is making an appearance. The technical costs are high. But in education, the human capability costs are higher (more expensive). The window to start AI plays in education is closing and we’ve done little from within the system itself.
This is good. How did we get here with AI? The Recipe for an AI Revolution: How ImageNet, AlexNet and GPUs Changed AI Forever
Towards Responsible Development of Generative AI for Education: An Evaluation-Driven Approach. Google paper on their use of supervised fine-tuning to improve engagements in education settings (in contrast with Gemini). It’s not a particularly impressive report on AI in general. But in education, it’s one of the better papers I’ve seen of the value of educational data in shaping LLM value to specific sectors. Google has been making sector-specific LLMs a focus, including healthcare.
What is higher ed’s role in providing AI skills? “We’ll need to be deliberate in how we educate and train our workforce, and how we put AI tools to use. Not only will higher education need to train the workers of today and tomorrow, but they’ll have to do so at the same time AI is transforming the daily operations and business practices of education itself.”
AI’s impact on jobs. “AI is the Industrial Revolution of knowledge work, transforming how workers can use information, find insights, and deliver results at speed and scale…Sectors that are especially exposed to AI are experiencing nearly 5 times higher growth in labour productivity…”Employer demand for many skill sets that AI can assist with to some degree — such as coding in Javascript — is declining fast, while demand for many skills that make use of AI — or are hard for AI to do (like sports coaching or ecological restoration) — is booming.”
AI in General
After Google and OpenAI had their big AI announcements last week, Microsoft stepped up yesterday: “The new features will include Windows Recall, giving the AI assistant what Microsoft describes as “photographic memory” of a person’s virtual activity. Microsoft promises to protect users’ privacy by giving them the option to filter out what they don’t want tracked, and keeping the tracking on the device.” Short demo. It’s already being investigated. The only thing more fun than watching AI advancements will be watching regulatory responses.
Everyone needs data. Slack wants to train models on users messages. OpenAI has partnered with Reddit for data. And with News Corp (Fox).
This is rather crazy: “in February that during the elaborate scam the employee, a finance worker, was duped into attending a video call with people he believed were the chief financial officer and other members of staff, but all of whom turned out to be deepfake re-creations.” $25m later… You need an organization safe word. Deep fakes are next level and getting better daily.
Saudi Arabia is investing $100b to become a global AI hub.