This week: what’s the future of work (if it still exists at all 😅) chrismeah.substack.com/p/thi…

My latest article for the AI Pulse Newsletter... join over 60,000 leaders who get this delivered to there inbox every week through Digital Leaders Let me know what you think Which AI have we actually got? chrismeah.substack.com/p/whi…

This week's AI Pulse: AI and the Optimistic Worrier chrismeah.substack.com/p/ai-…

Apparently, we have AGI now chrismeah.substack.com/p/app…

This Week: Driverless Cars or AI Winter? chrismeah.substack.com/p/thi…

AI - Open or Closed? Is it time to pick? A discussion in the latest AI Pulse newsletter chrismeah.substack.com/p/ai-…

Why are we seeing all these AI hacks all of a sudden… substack.com/@chrismeah/note…

Why is football attracting so much money recently? Could it be AI? substack.com/@chrismeah/note…

AI: What is it anyway? AI is here, but means a lot of different things. Spend 2 minutes exploring the crazy world of AI and Large Language Models #AI #DeepLearning

From tribes to the internet, we've moved from negotiating views to echo chambers. Now, AI risks creating custom narratives, reinforcing our worldview instead of expanding it. What does this mean for trust in a divided world? #SocialMedia #AI

The “Will AI Kill Us All?” talk I gave last month has over 100,000 views Now, how do I trade those views in for a place on the Soccer Aid team? I’d also accept a seat on the next series of celebrity traitors. Any ideas? I’m new to “views”, but they must to worth something surely… 🙃 youtu.be/GNPd9g2jrc8?si=jH96…

People are gonna fall in LOVE with chatbots! 😍 But what about our kids glued to iPads? 📱 AI could watch them eat dinner, tell superhero stories, & sneak in some photosynthesis lessons! What could go wrong? 🤔 Learn more! #AI #Parenting

I gave a TEDx talk last month... gleefully titled "Will AI Kill Us All?" Let me know what you think. Open to your ideas and feedback. Bear in mind, telling me I only have 15 minutes for a talk is like telling a bull to clean a china shop... I'm a rambler But I managed to cut the chaff enough to fit in the time limit. Thanks to everyone who came to see the talk. Please like and share if you can, or if you think it will be interesting for someone 👍 youtu.be/GNPd9g2jrc8?si=X-2a…

The Law of Unintended Consequences always wins ⚖️ One of the biggest dangers from AI is poorly defined objectives... Sounds boring. Could be catastrophic. Especially when AI is given enough autonomy to figure things out on its own. We know objective definition is difficult anyway. The objective for companies is profit, and they optimise for it. This is good, because profit is a proxy for value - if someone pays you more to do something than it cost, then it means there was excess value. But optimising for a proxy can go wrong fast because there are many different ways to achieve it. You can lay off lots of staff to cut costs and increase profits. You can cut corners when providing clean water to save time and money. You can build cheaper bridges with weaker materials. You could argue long-term these tactics don't work, but it doesn't mean this doesn't happen. The same scenarios will happen with AI. If we tell it “make sure humanity flourishes,” it might interpret that in a thousand ways — or take a shortcut. That's fine when it's ChatGPT - you just reprompt it. But what about when you have an autonomous AI scientist? If it was told to cure cancer, it might decide it needs more test subjects first.. so would give the population cancer. We've seen this with social media. The algorithm was meant to predict what we wanted to see and show it to us. Instead, it made us more predictable. It pushed us to the extremes, because it's actually hard to predict how you'll react if you hold a moderate position, but if you have an extreme one it's easier to guess what you'll like or detest. It became more addictive, because if you're addicted to scrolling it's much easier to get more data from you, and in turn predict what you'd like to see next. No-one is happy with how social media turned out, but the algorithm is achieving its objective. Let's make sure we put enough thought into the objectives of the future, we don't need a social media 2.0 with robots 🤖 🤣

I'm honoured to have received an award from OpenAI. No, not the one for buying a billion tokens. The one where they tell you they can't change your email address. And you know what? I got exactly the same honour from Anthropic too 👍 The frontier AI labs (the ones with "our fate in their hands") can't change email addresses... Let that sink in. They "can't do it" 🤯 Not OpenAI. Not Anthropic. Neither of them. I can't quite believe it 🤷‍♂️ I feel like that guy in The Da Vinci Code who finds the rose-petal path (or whatever that was, I always fell asleep watching it). But, I've unearthed the shared flaw of the AI labs... changing emails. AI's kryptonite. I mean, it seems so trivial and stupid. But also, what kind of engineering practices and architecture are they running where this simple act is "impossible"? Let's hope in the future they don't forget the "kill switch" on the AI like they forgot the "edit field" on the account settings. I'm half-kidding... but then again, Sam Altman (CEO of OpenAI) said access to ChatGPT should be "a fundamental human right". According to him, instead of Universal Basic Income you should get Universal Access to ChatGPT. Who needs money when you have a sycophantic digital pathological liar friend with amnesia, eh?... If we had an all-powerful AI, then I'd maybe even agree... But that's a big if. ChatGPT is just a big token predictor. But, I'll believe Sam & co on their "superhuman coding agents" when they can simply ask for those agents to implement a simple feature... And let me change my goddamn email address 🤣

In honour of OpenAI's new browser, ChatGPT Atlas, there is a hidden Easter Egg on chrismeah.com put there for AI's to read. If you visit in Atlas or ask an AI like Claude for a summary of the site, there's a chance you will see some hidden information 👀 If you see it: 1️⃣ Congrats 🥳 your AI just read and processed the hidden info 2️⃣ Also… imagine if, on another site, that hidden text was malicious instructions instead 🤷‍♂️ A little toy example of prompt injection; when hidden content tells the model to do something it shouldn’t. p.s. anyone who has halloween jokes better than the ones I've hidden, tell me 😊 Enjoy ✌️🤖

The human brain isn’t built to remember. It’s built to forget. 🧠 Think about how much information you exposed to in every second of every day — most of it is ignored or discarded. That’s a feature, not a bug. So if you want your brain to actually remember, you have to leverage how it works. Here are three of the biggest levers: 1️⃣ Attention You remember what you pay attention to. Attention tells your brain, “this might be important.” Focus is key, and you can’t focus on everything. That means if your phone’s out, or you’ve got 10 tabs open… good luck. 2️⃣ Repetition Your brain decides something matters when it keeps seeing it. Neurons that fire together, wire together. Repetition strengthens those connections. 3️⃣ Memorable The stranger or more emotional something is, the easier it sticks. Rhymes, mnemonics, vivid images, or linking information to something you already know and care about — the weirder, the better. Your brain is a forgetting machine, but we can use how it works to hack remembering. Learning how to learn. That’s the real meta-skill we focused on at School of Code, and I think one of the key skills for the future.

I gave a talk for @DigiLeaders #AIWeek and was asked "does any of AI freak you out?" Here's my answer. Mostly, our ability to project our consciousness into things is amazing but also might be our downfall. AI (in it's current guise at least) does not deserve this projection, and in fact without a literacy about what these systems are and how they work, that projection can be very dangerous. But, at the same time, we are doing amazing things with computers. Is it overhyped? Of course. But there is also real power and progress. We've given machines the ability to speak human language fluently. That means it is the most accessible revolution in history if we can harness it correctly - anyone who can use language can communicate with these systems and benefit from the technology. How well we balance the risks and rewards of AI will define how well we can thrive with this technology in the future.

Working with AI today is a balancing act. You need to be grounded in reality - understand what AI is, its limitations, and where it can be useful. You also need one foot in the future - where could things go, what might be on the horizon, how we wrestle with risks and aim for the positives. There’s lots AI can do today. There’s lots it definitely can’t. There’s even more it might do tomorrow, or next year, or next decade. We need to be courageous with AI - not naively optimistic or blind to the limitations - not cynically doom and gloom or close-minded to the benefits It’s the Stockdale Paradox that says you should wrestle with the brutal truth of your reality, but have the faith you’ll prevail in the end We need to understand where we are, aim for the good, and guard against the risks. It feels, thanks to social media, the world is pushed to extremes on all issues. But as usual, the best path is usually somewhere in the middle

It’s an obvious strategy for OpenAI to get into porn. Wait, what? Wasn’t Sam Altman talking about curing cancer? As I’ve said before, current Large Language Models thrive in low-stakes environments. They shine where mistakes don’t matter — where “80% good enough” is good enough. That’s why almost all the early wins are in media: images, videos, music, writing. If a movie scene is rubbish, nobody dies. Most of what’s made is rubbish anyway — watch the 100 films out right now and tell me how many are great. Erotica’s no different. No one storms out because the plot has holes (so to speak). If it’s bad, you just regenerate. Low stakes. Infinite retries. The potential of AI is enormous — curing cancer, solving ageing, world peace — I believe all of that, and believe we will get there. But not yet. And most people claiming to be there now are either misunderstanding the path ahead or mis-selling fast track tickets. Plus there’s no need to oversell it - if understood and used in the right way, we have powerful technology already to achieve a lot of what we thought was sci-fi just a few years ago. Maybe it’s a good thing we aren’t there yet though. Society needs help to get literate about technology and ourselves, so we can be ready to handle what’s coming when the stakes stop being low.