Good AI, Bad AI
While the potential AI offers is immense and can result in tremendous good for humankind, AI can also be misused.
We've just adopted a pup, an extremely friendly, playful and quite naughty Labrador Retriever who's perhaps around six-seven months old. He was a saint on the day he came home last week, perhaps anxious in a new environment. But now that he's comfortable at our place, we are experiencing the not-so-desirable side effects of a pup at home.
Yesterday, some indoor plants got a drastic pruning that was completely unnecessary, and every day he finds new possibilities for mischief. He's also learnt that he can grab stuff and bolt and will be given a treat to make him drop whatever it is he has grabbed. Long story short, since the past week, the most common terms you can hear at our home are "Good Dog" and “Bad Dog.”
That reminded me of ‘Good AI’ and ‘Bad AI’. While the potential AI offers is immense and can result in tremendous good for humankind, AI can also be misused. Last week, we looked at how AI can be used to solve India's massive pollution crisis, and if the authorities are listening and actually manage to do even a little of what our authors suggested, we could end up saving many lives. Because, as we are all too painfully aware now, pollution kills. This was a perfect example of ‘Good AI’.
‘Bad AI’ on the other hand isn't just a scenario of machines taking over and ruling over all humans; we'll leave that to Hollywood. ‘Bad AI’ is actually about the power of AI being misused. Some examples could include illegal surveillance, misuse of data, AI used for cyber warfare to bring down public infrastructure such as power grids; misleading medical advice given by a chatbot which could harm someone, and in one memorable example, a chatbot in New York rolled out by the local administration, which suggested business owners steal their workers’ tips or pay less than minimum wage. Much of it comes down to training AI and setting up the right guardrails to ensure AI implementations don't go rogue. As part of our coverage on AI, we will cover both good and bad, so that you know how to get AI right and avoid issues with AI that could cause you reputational loss or worse.
In the past week, we covered an interesting conversation between Yann LeCun, chief AI scientist at Meta and considered one of the godfathers of AI, and Nikhil Kamath. LeCun believes that the key to solving many of the world's problems lies in making people smarter and says AI can be a catalyst for that transformation. They also dealt with the ethical aspects of AI, a reminder of how important the Good AI/Bad AI topic is.
In another example of ‘Good AI’ at work, IIT Bombay professors have developed a solution that calculates the temperature of lake water surfaces considering the dynamic changes in its areas, which helps track the ominous march of climate change and will provide an early warning system for lake ecosystems under environmental stress, thus ensuring we can do something before it’s too late.
One term you will hear more often in the coming years is Agentic AI. While we have witnessed AI generating content and images and making predictions, we will now see AI agents capable of not just performing complex tasks but taking decisions autonomously. If you thought your bank chatbot was useful in some cases but didn't really seem "intelligent”, wait till next year, when you will see agentic AI at work in some of these customer service chatbots. Prepare to be amazed.
We have a detailed story based on insights from John Roese, global chief technology officer and chief AI officer at Dell Technologies, on what 2025 will hold for AI and agentic AI takes centre stage. It's a must-read.
Till next time,
Ivor
