Meta has poached Apple’s head of AI models. Techcrunch.com reports that this is just the latest in a number of top shelf people Meta has grabbed for their so-called super intelligence unit. Ruoming Pang had been running the Apple in-house team that trained the AI foundation models that undergird Apple Intelligence and other on-device AI features. Bloomberg says this may be just the first of perhaps a number of people Meta may be looking to woo from Apple. Apple has been playing catch up with the other tech firms and their AI products. They are far behind OpenAI, Anthropic, Google, and Meta. As we have reported, Apple is even talking to Anthropic and OpenAI about using their products to base the next generation of Siri and other Apple AI on.
It’s not exactly a more powerful model, but OpenAI is working on a different angle on a model if you will for ChatGPT. According to bgr.com, they have something new called ChatGPT ‘Study Together’. It has already started rolling out to some ChatGPT Plus users. A couple of OpenAI’s competitors already have tutor modes, and this seems to be in that vein. Instead of a student asking it questions, this model will allegedly ask questions that guide students to find answers themselves. The idea is to get people to think more, not just use the large language model to answer questions or solve homework problems. No word yet on when it will be generally released or how it might be priced.
Bluesky has been slowly adding features. One that has been missing until now, that was available on old Twitter and other sites was the ability to turn on notifications for specific accounts. Now, with Activity Notifications, that feature is live. Theverge.com notes that you can simply hit the bell icon on a page you want to get notifications from, and away you go. It works with news sites if that’s your thing, or you can use it to be notified of friends’ posts, too. Another addition is that you can set Bluesky to notify people if someone likes or reposts something they have posted.
Researchers at the University of Washington and Microsoft Research have used machine learning to develop a novel solution for trapping carbon in concrete by blending a sustainable, easy-to-grow green seaweed into the industrial batter that makes concrete, all without reducing its strength. Geekwire.com reports that the process lowers the cement’s global warming impact by 21%. Concrete in and of itself isn’t an issue, but the making of it produces some 8-11% of global carbon emissions. Scientists have been trying to curb its carbon footprint by using clean energy to generate the heat needed to produce it, and by swapping different ingredients. The product here uses dehydrated seaweed to make a high performing, lower carbon concrete. As a friend who was in that business often reminded me, cement is an ingredient of concrete…although lots of us who are laypersons tend to use the terms interchangeably.
I’m Clark Reid and you’re ‘Technified’ for now.