Duncan Anderson’s Post

The Future of AI Agents Who are /dev/agents startup? Towards a New Computing Paradigm Former Android VP Hugo Barra says AI needs ‘an Android-like moment.’ Are AI agents the real deal? AI companies are pushing AI agents as the next big leap in AI tools, promising digital assistants that can carry out tasks and make decisions autonomously and with little human input. But will they? /dev/agents is now out of stealth. https://lnkd.in/gVzNdMQr

Decentralized AI is finally getting mainstream recognition 🛫 When you consider the rugpull that Altman & Co are doing at OpenAI , decentralized AI becomes even more important

The AI hype cycle has dominated the startup ecosystem. The hype was big. Huge. Even though it feels like we’ll eventually hit the trough of disillusionment, I keep hearing about more useful AI applications.  We're at this crossroads, with two trends converging: 1. The hype energy running out (it's already starting) 2. The actual utility of AI increasing Which line intersects first? Does the hype live up to itself before the energy fizzles out? We might need a GPT-5 or some other big breakthrough to push us through.

What's hot in AI? - I blogged the other day about what’s hot and what’s not in fintech, but what about what’s hot and what’s not in AI? Equally, as I also blogged, AI is not one thing. It’s a million use cases which are all different. So, what are hot use cases for AI? Well, luckily, Allie Garfinle https://lnkd.in/eMxrxE_2

We All Know AI Can't Code, Right? -- The people who know how AI works need to speak up https://lnkd.in/e6b4xcpz #technology #startup #innovation #ai

AI is NOT overhyped... Is there a lot of noise and hand-waving about AI? No doubt. Are most generative AI startups doomed to fail? Certainly. Is the industry replete with snake oil? I'd bet on it. But even still, I believe that the world just isn't ready for what's coming. I think most people have a hard time parsing out the bandwagon, gold rush, "everything is AI" behavior of most companies -- from the economic seismic shift these technologies are building towards. And even worse, most people (including myself) are severely overestimating the time horizons of some of these shifts. About 6 months ago, I started a weekly AI call with a group of close friends to work on prototypes and share news. During our very first discussion, I remember distinctly saying something like: "Oh yeah, AI is going to change everything... But video generation is going to take at least 2 years to be usable professionally." 6 months later, I'm eating my hat. First Sora, now this...

What's hot in AI? - I blogged the other day about what’s hot and what’s not in fintech, but what about what’s hot and what’s not in AI? Equally, as I also blogged, AI is not one thing. It’s a million use cases which are all different. So, what are hot use cases for AI? Well, luckily, Allie Garfinle https://lnkd.in/eEsHHwdA

The most interesting thing in tech: is it possible to train a competitive LLM in an entirely decentralized way. A new startup is trying that and has just released a 10b parameter model. If this works, it could make AI more democratic; it could open up competition; it could let many more parts of the world collaborate on building AI models. And it could also make it much harder for anyone to regulate or control how AI develops.

Is your startup using AI? We're not an AI startup, if that's what you mean. I'm not developing a new LLM or a processor which more effectively handles AI related calculations. But are we using AI? We would be stupid not to. There are tools which speed up development, tools which help with ideation, tools that make things look better. It's also very probable that AI will find a place in the product we offer. There are a number of approaches which come to mind. But I'm not really thinking of those as AI, I'm just thinking of them as 'yet another library or API'. Talking about an AI revolution is real... but it's not playing a part in building the infrastructure of the revolution that interests me: it's how the outputs of the revolution increase my ability to get things done. At the moment these are still being conflated. And we need to understand that things like prompt engineering are both valuable (in terms of unlocking the capabilities of our new tools), but in the long term will seem more like 'having the ability to program EDSAC' - interesting, but not useful. Meanwhile my approach to AI will be: Continue to play with the cutting edge in order to see where it goes. Use it to make my life easier Build uses of AI into my code - but wrap them up so they can be easily replaced as things improve and standardise.

We All Know AI Can't Code, Right? -- The people who know how AI works need to speak up https://lnkd.in/eUzYWrGk #technology #startup #innovation #ai