AI and the need to refocus

The past few months have been an absolute rollercoaster with AI announcements, from incredible new models and conversations held with AI tools; to weary, doomsday-type news about the dangers of these systems. We have seen it all.?

We have also seen this before. AI has had multiple waves, and yes, this one may be bigger and stronger, but its technology and progress were not particularly unexpected.

So here are a few points I have been debating over the past few weeks:

1. Our jobs are (not) safe

Jobs requiring critical thinking, reasoning, contact with people, and decision-making are not likely to be passed on to a machine soon. Much like we have trouble trusting machines to perform surgeries, companies need a human decision threshold, and this can be a good thing to keep accountability and ethical reasoning.

In some ways, we have spent the last few decades stifling critical reasoning and fostering a culture of following and a need for belonging to a community that thinks alike - maybe it is time to shift the focus to promoting opposing views, dissenting arguments, and a culture of dialogue that seems to be lost. The less we behave like machines, the less they are likely to replace us.

2. We need education

Probably my favourite topic. We absolutely need education. We cannot expect people to work with AI if they don't understand it, and the fraction of the population that truly understands these systems is still slim. Some of AI’s dangers come from being misused by people who do not understand it, but we also need more people who understand its basic function to spot misuse. Think of it this way, we all have a basic understanding of maths - so we wouldn't believe anyone that tells us that a 10% increase on a 50k salary means earning 52k.

However, this is a complicated topic, and the extent to which everyone needs technical understanding isn't clear to me. But basic principles of machine learning have to form part of curriculums from a young age.

3. AI can be dangerous

Absolutely it can. If misused, the dangers are plenty. In my opinion, we are entering an era which needs that absolute alignment on ground rules for AI - like a UN charter. Funnily enough, Isaac Asimov created some in the 40s that you may have heard of (do not harm humans, obey orders, do not harm yourself), and that seemed like a distant future need. That future is now, and we need to ensure we implement them correctly (a lot can be discussed about these laws and the consequences of their application, but I mean this as a thought experiment and not a full endorsement in their original form).

We need transparent, aligned, ethical AI systems that we can understand and trust, to mitigate the risks of increasingly autonomous systems.

?

4. We need regulation

Yes, yes, and yes again. But we need regulation that works WITH technology and not against it. Regulation is slow, and technology is fast. But we need to have regulatory provisions to ensure AI systems are developed ethically and responsibly, and that there is a duty of care by those who are developing them. Data Privacy was a great example of this a few years back when GDPR, even with the raft of problems EU legislation may carry, has fundamentally changed the way companies approach personal information - even if, in most cases, their use cases haven't changed much; more importantly, it has changed the conversation around privacy and what it means, and that can only be a positive.

??

5. We should not halt development - but we should refocus it

Although I don’t fully disagree with last week’s open letter, signed by some of the brightest minds in this industry, I find it impractical and probably unenforceable. Furthermore, rather than halt innovation, I believe we have to refocus it from simply pursuing larger and more powerful models - almost like a Cold War-style arms race - to developing technology while improving ethics, algorithmic alignment, responsible development, policy and education. We probably need more development in this area, and not less - as long as we incentivise doing the right thing.

One thing is for certain. The latest AI technology wave has sparked conversations we should have had for years - which is definitely a positive outcome.