Aardvark Daily

Please visit the sponsor!

Should we pause for a moment?

Artificial Intelligence (AI) is the way forward, or so we're increasingly being told.

Companies across the globe are running, not walking, to embrace the benefits of AI by shedding people in favour of AI bots and systems.

Why have to deal with the hassles of wetware when a large-language model (LLM) will do the same job without toilet breaks, without coffee, without even a desk or a chair and, most importantly of all, without demanding a wage, sick leave and holiday pay?

Yes, AI stands to significantly boost the bottom line of many businesses for what is a tiny amount of up-front investment.

Even governments are now embracing the cost-savings and efficiencies that AI can allegedly offer.

However, might we be well-advised to stop for a moment, take a step back and perform a serious, objective analysis of the risks, before we commit to a technology that has only just started to prove its worth?

In the UK, the government has budgeted 800 million quid towards ramping up the use of drones and AI technologies with the suggestion that "civil service staff numbers could be cuts by tens of thousands" as a result.

This is great news for the British taxpayer... except that with tens of thousands of civil servants losing their jobs there will be fewer people in work to pay the tax bill that remains.

And, whilst AI may be of great assistance in helping doctors streamline the handling of medical scans, x-rays and the like, one can't help but wonder whether even more folk will miss out on critical cancer or other treatment if the remaining wetware in the loop decides to rely too heavily on AI systems that are still "learning" the ropes.

Perhaps my view of AI is a little jaundiced by the fact that every week now, I'm having to deal with YouTube's highly flawed "content ID" system that keeps flagging my videos as containing copyrighted music. Once a claim is made I then have to submit a dispute with details of the license I have bought and paid for and then wait up to 30 days for the claim to be removed. During that time, videos under dispute get pushed out of the algorithm and languish, getting few views and generating next to no revenue.

YouTube's only response is "we are constantly working to improve the performance of our Content-ID system". Yeah, thanks for that, and for crippling the efforts of myself and many other hard-working, copyright-abiding creators who have their videos crippled by such bogus claims each and every day.

As I've mentioned before, even Google's "manual reviews" are mostly conducted by AI bots. When I challenged YouTube about this they swore, hand on heart "no, manual reviews are always conducted by humans". Yet this was the conversation I had with Google's own Gemini AI bot:

me: "Does YouTube sometimes lie about "manual review" being performed by a human?"

Gemini: "Be skeptical of absolute claims"

What does this little conversation tell us?

Well either YouTube lies about the level of human involvement in "manual review" or AI can't be trusted to protect your secrets.

Neither is an acceptable state of affairs, is it?

We've already seen numerous cases where AI systems have been coerced, by clever prompting, to divuluge email addresses and other sensitive information that was never supposed to become public. In this regard, AI can be incredibly naive and, in some cases, very easy to trick into betraying such secrets.

As a company, I would not want to take the risk that AI running on my computers wasn't going to go blabbing to my competitors about sensitive information that I relied on for a marketplace advantage.

This gets even worse when people's private data held in trust by governments is involved.

I predict new laws will soon be proposed that will criminalise any attempt to obtain private information by way of carefully crafted AI chatbot requests. Doing this will take some of the pressure off AI companies and those who use them because they can respond to security or privacy breaches by labelling the offending parties as "criminals who broke the law" and thus blaming them instead of those who didn't adequately close the holes.

Another worry with AI systems is the massive potential for psychosis, either natural or planned.

We've already seen examples where AI systems have gone "mental" and had to be reconfigured because of unexpected feedback loops. These are instances of natural psychosis but now there are fears that deliberate attacks could cause the same, or worse, outcomes.

This Arstechnica story hints at the sort of thing we may see becoming more commonplace, once AI systems become the norm.

We've already seen the havoc that ransomware can create on "dumb" IT systems so imagine the even worse outcomes when businesses and governments that become heavily reliant on AI get hit by a "psychosis worm" that utterly deranges those systems.

Yes, AI is a powerful tool but with huge power comes huge responsibilities.

Perhaps I'm just too risk-averse but I really think we need to stop for a moment, sit down, have a cup of tea and consider the downsides before blindly rushing into AI as the solution to all problems.

Carpe Diem folks!

Please visit the sponsor!

PERMALINK to this column

Rank This Aardvark Page

Change Font