Only 10% or so of iPhones currently in use can even run AI on device. I think Apple is happy using OpenAI for now. They didn’t even mention adding Gemini as an option at WWDC. Or did they?
AFAIK Microsoft and Google have rolled out AI to all of their business/paid users, and OpenAI is raising money to build their own data centers.
Elon Musk now has 200,000 Nvidia chips in his “Colossus” computer in Memphis (and around 100 of his employees are still working with DOGE, which has access to all our Social Security data ). It is supposed to help get him to Mars. I have no idea what else he will do with it.
Apple has been using “machine learning” for years to improve their hardware and software. I think Apple Intelligence is mainly about replacing Siri with an AI Personal Assistant on the iPhone and iPad.
The rather pessimistic view that I’ve unfortunately ascribed to is the notion that Technology is indeed wonderful but when the choice is to propagate new technology as the expense of revenue/profit. Companies do not have a problem with intelligence when it doesn’t subvert their economic interest.
A few years back I was working in the Solar industry and I saw the conundrum monthly. States that had significant fossil fuels generating revenue were openly hostile to anything solar. As one tweeted “only Capitalism could make endless clean energy a problem”
This is a microcosm if every sector. AI should be a boon the the political process. Feed a proposed Bill in the system and AI with the right prompts should be able to work wonders. The realist knows Politics is intertwined with money and it’ll take more than a good AI game plan to affect large change.
This issue is always going to come down to the human element and what we are disciplined enough to engage in.
And neither currently exists. But given that AGI is nonexistent, it seems ridiculous to be talking about pursuing anything after it. You can’t build the second floor of a house without having a first floor.
Indeed! But, Meta seems committed to spending billions in the pursuit of superintelligence. Of course, they also spent billions in the pursuit of the metaverse. I haven’t seen it yet.
While I do not agree with Searle’s overall conclusion, I do believe his Chinese Room argument applies to attempting to achieve actual intelligence with LLMs. And is an instructive read for anyone interested in AI.
LLMs are a breakthrough technology and do have their uses. And they are likely the first processing module in an engineered intelligent system. But true understanding will not come via statistical analysis of word patterns. A symbolic processor is needed and probably a way to interact with the physical world as well. If Meta, or others, are spending their money in symbolic processing research then it may be money well spent. If on the other hand it is just toward larger LLMs then it is money wasted.
The UK government is committing £billions to “AI” both to run the government and to establish an economic lead in a vital new area by making the UK highly attractive to people like Open AI.
I’m seeing an increasing number of articles, based on an increasing number of academic papers, which raise doubts about the wisdom of this based on current technologies, even from people who are sympathetic to the general commitment and to the government. e.g. Today in the Guardian:
