Dr. Paul Sant, Dean of Computer Science at the School of Law, discusses the importance of developing responsible AI amidst cybersecurity challenges.
There's no denying that ever since OpenAI announced ChatGPT and we became aware of generative AI, there has been intense scrutiny of what's possible. The number of products in the artificial intelligence (AI) ecosystem has also increased significantly.
We have witnessed remarkable advances and innovative technologies emerging from the field of AI.
However, this progress has a dark side. What happens if an AI tool develops “thinking of its own” and performs actions we don’t intend? What if an AI outperforms its programmed capabilities, or worse, if a cyber attacker uses it for malicious purposes? What if it is misused for a certain purpose?
These are not mere speculations, but potential risks that require our attention.
Challenges in developing responsible AI
All of this is possible, and there has certainly been a lot of work done behind the scenes, but should we really be worried? Won't AI engineers realize this? Won't efforts by governments and institutions like the EU put in place policies and regulations to stop this?
Well, the answer varies. Indeed, developments like the EU AI Act set out a strong intention to ensure responsible AI development, and governments such as the UK and US have put together guidelines.
However, this needs to be balanced against the enormous potential that the transition to generative AI and general artificial intelligence is recognized to offer to capitalist societies.
So while there is certainly a desire to ensure responsible AI development, several conflicting perspectives make this more difficult.
AI impacts cybersecurity
To put things into perspective and show that there are people (i.e. cyber attackers) looking to exploit the power of AI for less desirable activities, consider the following: There are currently AI tools on the market that can collect data on small scales (less than 15). -Second) Create a sample of your voice and create an artificial intelligence version that sounds the same.
Don't you think that's “cool”? Well, that's rather “cool”.
However, consider that some organizations are moving to so-called two-factor (or in the event multi-factor) authentication to prevent cyber attackers from accessing things like bank accounts via online accounting. “Your voice is your password.”
Now imagine you've been interviewed somewhere and it ends up on YouTube or there's a website that includes a video introduction of you speaking. What can you do to stop cyber attackers from taking that sample, extracting your voice, and feeding it to you? Will it be incorporated into AI tools that can generate audio on the fly?
The answer is “actually nothing.” Now imagine they call their online bank, play a sample when asked, and are provided access to their bank account. This is because the system that checks your password cannot tell if the AI sample is your “real” voice.
While this is an alarming example, rest assured that those in cybersecurity are working hard to combat such scenarios.
Researchers working with large language models also ensure that those who attempt to modify these models will have unintended harmful consequences. A lot of work is being done to ensure that responsible AI has the upper hand to keep us all safe.
Realizing the true potential of responsible AI
The challenge we face is that the technology world is changing so quickly that products are brought to market without fully understanding their broader implications.
Of course, that's exactly what cyber attackers want. That is, it does not take into account the possibility of a one-time mistake or unintended consequences.
But by being aware, paying attention, slowing down, and thinking before we act (as a user), we can reap all the benefits these exciting technologies have to offer without exposing ourselves to risks.
There are certainly exciting times ahead, and we hope that we can realize the true potential for responsible AI to win.
There are many cyber defenders working to keep us all safe, so keep an eye on this area.
