AI and the Future of HE -
Hi
Hope you had an excellent weekend.? As the flood waters from Typhoon Yagi slowly went down, Hanoi came to life again with the sound of people celebrating being outside in the sun again and, oddly, chainsaws - as the city worked through clearing the 40,000 trees that fell across the city over the course of the storm. Bit different but then it’s been a different kind of year… enough talk though - let's get into it!
The Dawn of Deliberate AI: How OpenAI’s ‘Strawberry’ Heralds a New Era of Machine Reasoning ??
OpenAI's latest marvel, the o1-preview model, dropped last week and they may have changed the world … again.? Known as "Strawberry," o1 marks a significant leap in AI capabilities, potentially foreshadowing the rumoured "GPT-5”.?
This o1 series represents a shift from rapid, intuitive responses (parallels with Daniel Kahneman's System 1 thinking) to a more deliberate, analytical approach (System 2 thinking).? By "thinking through" problems before responding, the o1-preview demonstrates unprecedented proficiency in tasks requiring complex reasoning, planning, and iteration. So these are tasks like coding, where it absolutely crushes previous benchmarks, or watch it “thinking” for nearly a minute as it works through investment strategies ??.
The implications of this advancement are far-reaching, blurring the line between artificial and human intelligence. As Donald Clark notes, we may be witnessing AI's mastery of critical thinking, a uniquely human domain, pushing us closer to AGI.? The model's ability to handle multi-stage problems suggests AI becoming an indispensable partner across various fields.? However, Ethan Mollick points out that this progress raises important questions about human-AI collaboration, echoing past AI milestones in chess and strategy games.? While experts suggest this breakthrough bodes well for accuracy, transparency, and ethical alignment, it's crucial to consider both the potential and ethical implications of machines that seem to "think." The University of Sydney's Kai Riemer cautions against anthropomorphising AI, reminding us that it's not actually thinking but rather reviewing outputs for consensus and consistency.?
Wherever you end on that question though, a final thought for now: they just dropped an AI that can “think” without fanfare on an ordinary Thursday in mid-September - what the hell do they have in store for us at OpenAI Dev Day in October?!?
AI Behind the Wheel: Is Waymo's Self-Driving Revolution 84% Safer than Humans?
Waymo, the autonomous vehicle company, has released groundbreaking safety data that could reshape our perception of self-driving technology and just who should be behind the wheel (spolier: ???> ??).? According to their latest report, Waymo's autonomous vehicles are 84% less likely to be involved in serious crashes (defined as crashes triggering airbag deployment) than human drivers.? This statistic, based on 22 million rider-only miles driven across multiple cities, paints a compelling picture of AI's potential in road safety.
The data doesn't stop there.? Waymo vehicles were involved in 73% fewer injury-causing crashes and 48% fewer police-reported incidents than their human counterparts.? Extra-interestingly, it appears that the large majority (~70%) of the crashes Waymo vehicles were involved in were caused by another car rear-ending a Waymo. These figures, drawn from operations in San Francisco and Phoenix, suggest that AI might not just be catching up to human drivers – it could be completely outclassing them in terms of safety. ?Of course, human oversight remains key, but these statistics hint at a future where AI literally takes the wheel - potentially saving countless lives in the process. ?On a personal note, as a motorbike driver in Hanoi - a city famous for traffic and … energising driving styles - I’m all for this ????
Move Over, Dr Bunsen Honeydew: AI's Cooking Up Some Wild New Research Ideas
OpenAI's o1 model may have just changed the game for general AI, but a new Stanford study is shaking things up in the scientific research world.? This study, involving over 100 NLP experts, found that AI-generated research ideas were judged as significantly more novel than those from human experts, though slightly less feasible.? Meanwhile, the folks at FutureHouse are adding fuel to this AI-powered fire.? Their system, PaperQA2, didn't just match human experts - it outperformed them in tasks like answering questions across scientific literature and producing factual summaries. And get this: it even spotted an average of 1.64 human-validated contradictions per paper in biology literature. Read Language agents achieve superhuman synthesis of scientific knowledge and ask yourself, is AI becoming the world's most thorough peer reviewer?
Super interesting stuff - but important to note that both studies highlighted some important limitations.? The AI-generated ideas often lacked diversity (apparently, AI hasn't mastered the art of "thinking outside the box" just yet), and struggled with reliable self-evaluation.? As AI capabilities in scientific domains continue to advance, addressing these limitations while leveraging AI's strengths will be crucial.?
领英推荐
Deepfakes in Academia: When Seeing Isn't Believing
Let's talk deepfakes. I'm no expert, but creating that picture of Biden and Trump kicking back enjoying the simpler things took about 30 seconds, tops - and that was almost a year ago (so ancient history in the AI world).?
From fake Biden robocalls to Taylor Swift deepfake pornography, this issue is intensifying.? The White House recently announced voluntary commitments from tech companies to combat AI-generated image-based abuse - a solid start.
In Higher Education, Dr Mike Perkins , Leon Furze , and Jasper Roe 's new study highlights urgent risks to academic integrity, institutional reputation, and student well-being. Deepfakes could be used for academic dishonesty, creating fake credentials, or even cyberbullying of students and faculty. Their research proposes a comprehensive agenda to address these challenges. They've launched a crucial survey on faculty attitudes - I strongly encourage sector professionals to spend 10 minutes shaping this vital conversation. As deepfakes threaten to erode trust in educational institutions, engaging with this issue is critical. The truth in our classrooms - our institutions - may depend on it.
AI-Powered Content Creation: A Revolution in Information Delivery and Learning
This was an absolute AI “wow” moment and I strongly recommend you give this a try ???
I used 谷歌 NotebookLM's Audio Overview feature to create a fantastic two-person AI discussion from a long-form, somewhat detailed text. Here's what happened:
1?? Fed the AI the latest “AI and the Future of HE” blog post
2?? Hit the "Deep dive conversation" button
3?? Sat back and listened in awe!
The AI hosts, let’s call them Dot and Dash, covered the last "AI and the Future of HE" post so that’s:
What blew me away: Their ability to contextualise complex info, use relevant anecdotes, crack jokes (?!?), and maintain a natural conversation flow. ?Don't take it from me though - listen to the podcast here. ?
Imagine the possibilities for personalised learning, multi-lingual content, and new ways to engage with educational research! This feels like the start of something very exciting ??
As we've seen, AI is rapidly advancing across multiple domains, from enhancing our cognitive capabilities to improving road safety and revolutionising scientific research. While these developments bring exciting possibilities, they also raise important questions about ethics, human-AI collaboration, and the future of various professions. As we navigate this AI-driven future, it's crucial that we remain engaged, critical, and open to the transformative potential of these technologies while also addressing the challenges they present. That's part of what we're hoping to do with this newsletter.
Have a great week ahead and let us know if there's anything we're missing that we should add to make this 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. ??