We have all seen the rise of AI-based technologies and the fear-mongering around the existential risk to humanity and how AI will take over all jobs.
Before I get into it – this article was NOT written by ChatGPT, Google Bard or any other LLM. These are my own words – but of course that’s probably what I would said if AI did write this blog.
Okay, back to the point, let’s put some cold water on that fire that AI is going to take your job.
Yes, AI will hugely revolutionise how we all work, how our employers interact and what they expect from us. But at the moment, AI cannot replace a human. Most of the improvement in AI in the last 12 months has come from Large Language Models, better known as LLMs, but their design models are built on learning from existing material fed into it from existing sources.
Human written and developed sources, from this data, can extrapolate, develop, modify and present a response to a natural language question from the data they have been taught, which needs to be closely monitored and checked to ensure integrity and verify its truthfulness.
One Big Flaw
This is where current AI draws its major flaw: it cannot distinguish fact from fiction – it only knows its data. This is why when you feed data back into AI from AI, the model degrades, and very quickly, you lose all sense of fact, and the results become conjecture or in some cases completely false. This is why, in the short term, AI cannot replace the human operating it because we all need to take the presented data and give it the once over and look at the text and say, “Does that really make sense, and is that really the truth or is it some biased opinion or even worse incorrect?”.
The same logic applies when you look at replacing human roles with AI; someone will need to be the gatekeeper to the operation, and someone will need to validate the output.
Some tasks are at risk
AI may well replace the roles that we currently spend hours on. For example, writing this article has consumed a good chunk of my time and is done by reading sources, digesting news and creating a written argument that pulls this together. In the future, AI will be the one writing this article, pulling from the latest data (This is important as current AI is using legacy data) to generate an article. This will then be checked over by myself and then a copywriter and published for you to consume. This won’t make the copywriter redundant, or me, but will change the expectations of our employer as to the expected deliverables of our roles. This is the key behind the headline that AI can’t replace humans in its current form. In future forms, it may well be able to alter many people’s working lives, but humans are still going to be a key part of the puzzle.
So how is it going to be regulated?
To come onto the second point around control and legislation, we are seeing a continued wave of posturing by Governments and Big Tech about AI and how it needs to be controlled. Some of this is playing to the crowd. With the general public so concerned about AI and how it will affect them, legislators are looking to make sure they have a stance on it and try to control the headline.
That piece aside, there is a very important decision to be made by Governments and the global community about AI: Who sets the boundaries?
Is it going to be Big Tech like it was with the Social Media Revolution, where the world changed, and Governments spent ten years getting up to speed on the technology and how to control and police it? Or will the Governments of the world take control? Will they try to limit the development of AI, stunting its potential with overly controlling legislation which prevents Big Tech and start-ups from developing with AI and pushing the boundaries of what we currently know AI is capable of?
What do I think?
Now, I see this as a very hard line to balance with some European countries’ swing for all-out AI bans before we have even got AI into our everyday workflows. It seems to be exactly what I mentioned before, stifling the development of those countries and the businesses that operate within them. That said, on the flip side, with no control and protection, we could see a world very quickly where AI not developed by us is being used to operate and control CCTV, Traffic light systems, and self-driving cars. This presents a very real risk to the citizens of a given country should the ability to control the AI fall into the wrong hands.
I believe that most regulation will come down to a risk vs reward model, where mainstream AI development is not prevented, but the integration of the technology too heavily into the core infrastructure or day-to-day safety of a country’s citizens will be restricted to prevent the risks I mention above. I don’t believe there is any silver bullet, and with any developing technology, you need to adapt and adjust to it. Businesses that are leveraging the technology need to be sensitive to the risks they present should their tool be leveraged by threat actors or hostile nations.
You shouldn’t fear AI in your business; you should also be aware that not all AI headlines are true to the nature of the risk. AI is amazing if used correctly, and if you ensure you understand the technology and put proper protection in place, it is a hugely powerful tool; when done incorrectly, you can easily put your business or your data at risk.
If you want to talk to one of our experts about how we can help you with your security and understanding of LLMs, then please call 01235 433900, or you can email [email protected], or if you would like to speak to me directly, you can reach out to me via DM or at [email protected].