#217: Gen AI - The Primacy of Language
Language is the baseline on which everything else works - Gen AI is taking over this level zero, to become the default intelligent communication layer between us and the world.
When little children first start going to pre-school, the parents are usually given one instruction. Talk to your child, often, all the time even. Engage. Communicate. Read to them. It’s not hard to see why. Communication is the foundation on which everything else is built. Once you can communicate, all other education can follow. Even maths or music education depends on our ability to communicate well.
When the 33 miners were trapped 700 meters underground in Copiapo in Chile, almost 15 years ago now, the most critical part of the rescue process was the ongoing communication that allowed information exchange, update on morale and rescue efforts, and allowed the miners to stay organised and focused for the rescue. A long pipe allowed messages to be passed back and forth and for daily communications.
Language and communication is fundamental. Everything stems from that. Our bodies work because the brain sends signals to our organs and limbs. We fall in love because of signals from the brain, and we feel hungry thanks to other signals. Language is a primary reason for the dominance of homo sapiens on the planet. Language is the social glue that creates communities. Language wires our brains and provides the building blocks for meaning. Language is the technology with which the humans were first able to transfer a thought from one person’s brain to another’s.
It’s in this context that we need to understand the critical role Generative AI can play in our lives, because it is able to synthesise and generate language. It can form that foundational layer on which all other intelligence can be built and delivered. If organisations are getting intoxicated with AI, then Generative AI is the gateway drug.
Generative AI as a default machine interface: Historically we’ve needed to teach people to use machines. Whether it’s a complex machine such as a car, or a ‘simpler’ machine such as a dishwasher. There is effort involved in learning the right sequences of buttons, keys, knobs, or pedals to push or turn. The primacy of language will mean that increasingly, we will only need to instruct the machine in natural human language. Note that this is distinct from autonomous machines. Even autonomous vehicles will need to be given instructions. Gen AI will allow you to say to your autonomous vehicles “take me home, via the supermarket, and play me the latest episode of Bear”
Generative AI as a programming interface: I just heard Jensen Huang, the CEO of Nvidia make this point. We can stop telling our children to study programming because in future, code will be written by AI, and all we need to do is bring deep domain expertise. We’re already seeing the early signs of AI driven software development. Many of my colleagues are working on these tools, with surprisingly good results. After all software development also uses languages with specific terms, connections, meaning, grammar, and syntax.
Generative AI will as a Default Data Interface: At the risk of repeating myself I am convinced that Gen AI will upend the world of reporting and transform analytics as we know it. You shouldn’t have to extract reports with hundreds of rows, and send to dozens of people so that head of sales can find the one piece of data she’s looking for about store performance last night, or the head of supply chain can see stock positions across stores. In the world of Gen AI, you should simply be able to ask the question, and if the data exists within your organisation you should be able to get an answer.
Generative AI as a default environment interface: It seems a pretty safe hypothesis that our environments will get smarter. Thanks to sensors, mixed reality models, and edge computing. Consequently the way we interact with our environment - our homes, classrooms, offices, or public spaces, will also evolve to a two way model. The smart environment will adapt to our needs (heating, cooling, lighting), whilst optimising energy utilisation, for example. And we will be able to engage with it in new ways, customising, personalising, and seeking more information and services from it. And all of this will be underpinned by a communication and interaction model which in the emergent world will be underpinned by Gen AI, which is another way of saying we will simply be able to talk to it.
Generative AI as a default human interface? This is the one that is probably the most contentious. I used to call my friends regularly, and before that I would call them on landlines. And even before that I’d shout from my window to my friend next door to see if he was available to play. Today I sometimes message them on WhatsApp to see if they’re free for a call. I text them to ask if they’re up for dinner sometime. Our conversations have become asynchronous and intermediated by machines. My daughter’s preferred way of connecting with her friends seems to be social media. It seems only a matter of time before Gen AI becomes the intelligent layer between humans. Perhaps not at home, or with friends in the first instance, but definitely at work, and through our various interactions. I’m not suggesting that this should happen, but that it probably will happen. Some of us will find this deeply disturbing. Future generations will probably find it normal. It may also lead to the point made eloquently, that rather than worry about the apocalypse, we should be more concerned about the death by boredom when your Gen AI writes a letter on your behalf to your bank and the bank’s Gen AI writes the response! Snapchat already has a suite of Gen AI based offerings.
Welcome to the world of instructional linguistics
What will this do to language? Will it improve instructional linguistics? Every major technology has had its impact on language. Most of us will remember txt spk, or the way in which internet memes have given us global lolz and the evolution of shorthand - IFKYK (which expands to if you know, you know, ICYMI!). The focus on prompt engineering might spill over into language in interesting ways. Will it mean that in future we walk into a coffee shop and say ‘Imagine you’re the best barista in the world, and make me a cappuccino, but reflect on the outcome and use the optimum amount of beans to wake me up’? Probably not, but nonetheless, it might surprise you to see the ways in which our interactions with Gen AI end up in everyday language!
AI Reading
Mistral - the French AI company backed by the government in France, is looking to rival OpenAI and build a European powerhouse, that according this piece uses politics as a key differentiating ingredient along with data and compute, for competing in the crowded AI and LLM space. (Economist)
Perplexity: It was always a matter of time before somebody did this. Perplexity looks to drive Gen AI into search, to improve on the traditional page ranking models. (IEEE)
Other Reading
Managing Innovation: I really like the phrase ‘collaborative advantage’ in this piece. And the line “innovation is more likely to succeed when it is a curated process in which an intermediary takes responsibility for sparking and sustaining collaboration among the people involved.” The piece talks about how to nurture innovation and makes some excellent points (HBR)
DORA Report: The State of DevOps report released by Google and DORA is illuminating and it also invokes the wonderful Goodhart’s law: “When a measure becomes a target it ceases to be a good measure”. So true! Here’s the report. It also makes some great observations about elite software development. (DORA)
BioTech: 3D printed prosthetic eyes are here. And they last longer than the current alternative for prosthetic eyes. (Nature)
Uber Robots: UberEats is using autonomous sidewalk robot for food delivery. (CNN)
Apple has reportedly scrapped its Project Titan - to build electric autonomous vehicles. The reporting on this is confused. Some people are calling it a sign of the slowdown of electric vehicles, which is mystifying to me. Others think it’s a sign that autonomous vehicles are taking too long. Or that Apple have struggled with the challenges of automotive design. To me, it seems like Apple has just decided to refocus on AI, just as many other large tech firms are doing and is streamlining attention to a much more near term revenue opportunity.
Have fun and see you soon.