Richard Matsui, co-founder of kWh Analytics, McKinsey China alum, and current advisor under President Biden to The U.S. Department of Energy’s Loan Programs Office (LPO), visited the iQ 360 team to share his thoughts about the current state of AI. He provided a framework for thinking about today’s AI applications.
Below, we share this framework, as well as our hot takes on where AI is headed.
At the moment, AI makes the most sense broken down into five categories:
1. Content Generation
- Write a novel in a weekend with ChatGPT
- Craft a customized bedtime story for your kid with Fable Wizard
- Convert a blog into a podcast in two minutes with Podcastle
- Dub any video into all languages with YouTube’s AI-powered dubbing
- Put anyone’s face on any video with Deepfakes Web
3. Autonomous Agents
- Provide calendar access and Google Duplex’s AI makes dining reservations for you
- Monitor global media mentions, identify negative mentions, reach out to the journalist with a rebuttal and create a daily coverage summary; all with AI
4. Invisible Patterns
- Analyze 1.5 million known drug molecules in three days and create a shortlist of 23 molecules with Equibind
- Confound experts and professional Go players with AlphaGo
- Pass the bar exam with ChatGPT-4
- Pass the medical licensing exam with ChatGPT
Next Up: AI Supertools
While the above framework works for now, we suspect that at some point, AI “supertools” will merge these capabilities together, similar to how our phone is also our camera, our calculator and our word processor (for those who remember word processors!) The various discrete tools will go away and will look primitive when compared side by side with what’s coming. Gaining a general understanding of current AI platforms will likely influence early adoption and understanding of future “supertools.”
AI Becomes AGI
OpenAI submitted a trademark application for GPT-5 to the US patent and trademark office in July 2023. Supposedly, this pending version is so powerful that it is poised to achieve artificial general intelligence (AGI). According to TechTarget, AGI is the representation of generalized human cognitive abilities in software so that, if faced with an unfamiliar task, the AGI system could find a solution. The intention of an AGI system is to perform any task that a human being is capable of.
Dissemination of Misinformation
Many highly respected newsrooms now block OpenAI’s web crawler, GPTBot, from accessing their platforms for content; among them are CNN, The New York Times and Reuters. Without reliable sources, future AI models may be left to train on misinformation and distorted reality, which would create a feedback loop of inaccurate content that will be nearly impossible to interrupt. From a reputation management standpoint, this has the potential to be a nightmare scenario unless we prioritize methods for validating information received from AI before utilizing it.
The promise of AI for invisible pattern recognition in such realms as healthcare and science is best encapsulated by Google’s AlphaGo beating the world’s best Go player. It wasn’t just that, after training for a matter of months, the software claimed victory in a game that humans have been studying for thousands of years. It was how it won that is so interesting.
“Move 37 goes against all conventional teaching and no experienced human player would ever have played it. In fact, we know from AlphaGo’s calculations there was just a one in ten thousand chance of a player selecting that move. It was a moment of inspiration that came from its unique approach to the game. Unlike the way I – and all other human players – approach Go, its decisions are unencumbered by the tradition, theory, and teaching of human play. Instead, it learns the game for itself, giving it the opportunity for fresh thinking and leading to a unique ‘free spirited’ style which in turn has unshackled human players from tradition and allowed us to also think differently about the game.”
- Fan Hui, professional Go player
AI is all about speed and scale. AI doesn’t need to sleep or eat or stretch its legs. An AI cancer research tool can hammer away at the same problem non-stop 24/7 until it hits upon the right gene, cell, or drug molecule. If healthcare outcomes will someday soon be dependent on access to AI tools via broadband internet, how should we think about the still unresolved digital divide afflicting too many rural communities? AI has the power to do so much good, but will failures of equity get in the way?
Header credit: Medium