I've talked about this on #TWIGAI and a couple of weeks ago I was talking about GenAI at the Hitachi Digital Services launch event and someone asked me if this whole thing of GenAI is sustainable. My answer was an immediate "no!" but "it's improving, we're coming up with better solutions to reduce the energy consumption and overall carbon footprint, but it needs to get worse before it gets better". I think we're just starting to untangle this new thing and it's not even yet for everyone. We don't have it on our phones (Apple Intelligence), there's nothing mass installed on the laptops (Copilot PC) and organisations are just starting to realise the benefits of GenAI. So there's a long way to go and obviously not having enough power to "power" the GenAI evolution and electric vehicles revolution, is a massive issue. The only solution is going to nuclear to power datacentres or clearly adopt renewables at a massive scale. Both depend on governments and people to accept them. Without it we can't have the "nice things" :)

We all heard that software is eating the world but this article looks at the historical development and challenges of software engineering, while Andreessen's view discusses the contemporary dominance of software in transforming industries. Software, at the end, is always about complexity and abstraction vs the economic and strategic benefits. But we can say that complex software always revolutionises the market and disrupts things. I'm thinking about all the LLM's launched for the past year, the complex algorithms and math being produced to cope with that new category. It's an amazing world!

Genuine art connects deeply with human experience and transcends time, offering a profound contrast to artifice, which aims to manipulate and control. But the commercialization of art in my opinion really dilutes its true essence and reduces it to mere commodity. Just like software development, that is currently driven by prioritising market features over robust, secure, and user-centric design. Software should remain useful and effective rather than becoming a mere product of commercial pressures, just like art :)

FCKGW-RHQQ2-YXRKT-8TG6W-2B7Q8 - Back in the days we knew that by heart, didn't we? Just like Windows was "freeware" back then, saying that if a content is on the web, then it's free, is quite a stretch... a weird flex one too, for the company that created the EULA. We clearly need to understand what content can and cannot be used (or how to remove it) from LLM's. Probably the same rule of the right to be forgotten on Google, needs to apply to what the LLM knows about us and how it generates content from us. Github Copilot has a metric of 65 lexerne - a basic abstract unit of meaning - that if the output code is matching something already available, it doesn't reply to you. The same metric should be used for LLM's where if the output is matching ipsis verbis something else, then it should notify us and reference that.

Ah!!... the nostalgic post! right? :) yes! Pretty much! It's a flashback of how crucial it was in the early video games to have character tilling due to the hardware constrains and limited memory. Today it feels that we can always add a better GPU and CPU or more RAM to it and don't care about limitations. But back in the days, that wasn't possible ($$$), at least for me that always had to support my gaming and earn for it. It's a great reference for today’s game developers.