In one of our internal annual leadership conferences at Intel, Diane Bryant, our then CIO, conducted a break-out session on out-of-the-box / lateral thinking. We were ~20 people in her session when she asked us all to write down the maximum number of use cases for a metal-wire hanger, you know the flimsy ones that you see in most motel/hotel rooms.
While most of us struggled with listing 3-5 use cases, couple of my peers managed to list ~10. The most common were – 1) cloth hanger (of course), 2) antenna and 3) car lock opener. She then shared that the same question was asked to elementary school children and they came up with 30 different use cases, on average!
This is exactly the situation with industry leaders who keep asking – what are the use cases for AI, or how can I use AI? Most have dabbled in co-pilot and chatbots but struggle to imagine anything beyond what is an extension of what the IT departments have been doing for the last few decades. We all seem to be stuck in the ‘fog of relevance’ of yesterday’s use cases of computing technologies for AI, which is anything but.
The creators of AI are not helping either, as they keep churning out copy-cat and one-up solutions – vibe coding, text, audio, image, and video generation – that has almost eliminated differentiation between them. After all, if the entire class has learnt from the same textbook (online content), taught by the same teacher (transformer architecture), and with very little differences in learning techniques (RL), what can you expect?
This is why I wish the AI bubble burst as quickly as possible so that everyone can stop drinking the same cool-aid and get a breather to come up with unique products and services, just like the real Internet boom happened after the dotcom bust.
In my earlier blogs I have shared some ideas for software products and services. Here I would like to add couple of hardware-software ideas. I would love to connect with someone who is already working on them, or I can help define the product strategy with whoever is willing to take them up.
AI in a box: Think of a small, fully contained, stand-alone AI inference device with incremental learning capabilities that can be plugged into your home router. This box becomes your household’s multifarious AI assistant, that will not just respond to your or your family members’, or devices on your home network, questions but act as a proactive AI agent to manage your household. Think of cybersecurity, dynamic bandwidth allocation, smart devices management, vacation responder, planner, budget management, and plethora of other tasks that you do to run your household effectively.
While OpenAI, Anthropic, Google, X and other AI creators would want to tightly couple their own models with their own hardware; chipmakers like Intel can create an ‘Intel inside’ box that would give the consumer the choice of whichever open-source model(s) they want to use or subscribe to.
Health monitoring: This industry segment gets me most excited about AI use cases. While the big pharma and FDA will continue to drag their feet, the consumer segment can, within bounds, help us lead a better, healthy life by not just tracking the fitness indicators but by using the AI to personalize the predictions and recommendations to my bio markers, routines, diet, and habits.
Imagine that my smart glasses watch me gobble down a double cheeseburger with 32 oz of IPA. It sends a notification to my ‘health AI’ that will watch for the spike in sugar levels and tells my watch or ring to nudge me to walk for 30 minutes after the meal is done, rather than just recording the event. After an hour if I reach into the pantry to search for a snack, the smart glasses can notify me ‘hey, dude… pipe it down, you had a 2000 calorie lunch an hour ago’. Over time, it can learn my dietary pattern, my lifestyle, my exercise pattern and, if I choose to share, my health records from annual checkup and more.
I just scratched the surface of these ideas here, but you get the gist. The enterprise will be slower to adopt AI than the consumers would, and there are a ton of ideas that can translate to meaningful products and services in that space. I wrote some of the ideas 25 years ago that can now be brought to life with compute, bandwidth, and AI we have available today.