The Future of AI
The Business, Geopolitical, & Societal Impacts
Preface
We’re in the third boom for AI. Systems are becoming easier to install and use. At the same time we’ve been told that AI is going to take all of our jobs. Yet AI often fails to impress and excite. Months on from rumours of Sentient AI nothing has come to fruition.
So what are we to believe about this Technology? It seems to have a breakthrough every other day, yet the results aren’t filtering down to the everyday consumer…
AI Safety
AI safety is a joke and does nothing, but harm model effectiveness. As discussed here a few months ago.
Hindering model effectiveness is quick path to a failed business and a wild and thriving Open Source underground of models. Which is what we see today.
AI like all digital technologies cannot be contained. Digital regulations are an impossibility; akin to Jailbreak on iphones, any security feature will be bypassed. The true idea of digital regulation is one of a cultural nature rather than an actual limitation.
Governments will try to regulate digital content and technology and will fail spectacularity. China’s Great Internet Firewall hasn’t stopped the Internet, it has just made access more difficult for many, and more valuable to those who have or want it.
The only way to regulate a digital technology is for it to be securely contained and compartmentalized, unknown to the public.
AI is already out of Pandora’s Box.
Nielsen’s Law: Users' bandwidth usage grows by 50% per year
Moore’s Law: Computer processing Power doubles every 18 months
AI Winter?
After every industry Boom has come a Winter. AI is not different, and given no one else is talking about it, I suspect we will be experiencing one soon.
First Moore’s Law is no longer active and is slowing down. Manufacturers are reaching the physical limitations of dye shrinkage at the 3 nm level.1 Elections at the 3nm and below begin to exhibit the effects of quantum tunnelling. Thus the only way to continue efficency and performance gains would be to stack chips. Meaning that future tech products have a size limitation.
Continued Chip shrinkage has lead to model inefficiency and has favoured throw more GPU horsepower at the problem to fix it. This idea of throwing more computing power at the problem is a result of continuing performance gains and cheap money. Neither of which will be around and will cause the industry as we know it to collapse. The Industry nor the idea will not die like the 60’s, 80’s or early 2000’s; but it will need to for us to gain the necessary industry changes and mentality to advance the industry and idea.
The torch will be carried by a few Businesses and the innovators hoping for a brighter future.
The Business Case (and Lack of) for AI:
VC money has been burning to fuel the growth of this technology to its success. Now it’s trying to find product market fit. Will they find a sustainable business model before server costs eat their runway?
Who knows.
But when selling a strict AI company product (like OpenAI), what will people pay for a chatbot? People may be paying 20 dollars a month, but that doesn’t cover the server costs. Especially now where someone can run a full model on their laptop without safety constraints.
We haven’t seen many successful businesses yet. Because everyone has pitched and drank the Kool-aid that AI can do everything. But it can’t.
In Reality AI has two functions that it performs well at: