Get Ready for Personal & Corporate AI

When IBM president Thomas J Watson was asked during the 1940s how many computers he thought the world would need, he is reported to have said: "I think there is a world market for about five computers."

Watson's legendary misjudgement did not prove fatal to his company. In some ways, he was right – we have a handful of big tech companies which are essentially giant integrated computers.

Most of the technology we end up with in our homes, or with personal access to, starts inside large institutions. The AIs such as ChatGPT, which are taking the world by storm at the moment, are in many ways ‘institutional technologies’ we are granted access to. But the pattern of where they will end up is clear. We’ll all have our own personal AIs. We’ll have them in the companies we work for and even on our own personal devices. This will be a huge shift resulting in differentiated revenue streams for big tech.

Computer Power

First, we need to understand how the AIs got to now. There are a few reasons why AIs have become so powerful recently. Firstly, we can thank the ever-increasing availability of computing power and cloud storage. These AIs need to be able to comb through extraordinary amounts of data, and so require massive processing power. ChatGPT has over 175 billion parameters it works through.

The other thing these AIs need is the data to train on. And that is where we come in.

In the past two decades we have populated the internet with the entire human experience. We have essentially replicated humanity and created a quasi-digital mirror of our species — everything we think, say and do. Think of all the video, our locations, the entertainment, the science, the knowledge, the images… it is a digital hive mind which these large language models feed on to create human like output. Human-like, because it replicates human input.

Teaching Children 

But this is just the start of where it will go. At the moment, we are continuing to train the AIs as we interact with them, which provides feedback loops. Every query and prompt we put in makes the AI smarter, as does our subsequent prompts. It’s a bit like teaching children. Once these large AIs are sufficiently trained on large datasets, they’ll be able to function on much smaller, even micro datasets such as those in corporate and domestic settings.

It isn’t just the internet we’ve been populating either. Our own devices are filled with everything that matters to us in our lives. We get a new device and transfer that data to the next and the next and so on. Most people have a complicated combination of data on their cloud storage and hard drives across their PCs and smart phones which is increasingly dense and difficult to navigate and retrieve data from. The data isn’t trivial either. All our pictures, our videos, our finances, our lives. This will be the next bastion of AI. Just like the computer did, the AI will become personal. We’ll all have our own Jarvis.

Searching, Creating

The main two functions will be very similar to what is happening in search engines where it is bifurcating into the two streams of either searching or creating. These personal AIs will eventuate and be let loose into our devices to crawl and find whatever we need. It will be far more advanced version of the search functions we have now with Finder or File Explorer. Anything we’ve ever created, which resides inside our personal dataset will be accurately retrievable based on inexact semantic search.

But the real boon will be the Personal Generative AI which can create whatever we need. It will do this by combing the training from open source AIs, with our own personal dataset. It will spit out work which has our style, personality and insights, our wisdom converted into written documents, to pitch decks, financials, anything. Just like we are starting to do now on the open web, we’ll be doing that on our own work history.

This will also be a massive unlock inside large corporations. One of the key challenges most large firms face is knowledge leakage. Every firm backs up everything it does every day, yet very little of it is findable. It’s even harder to uncover and correlate data across divisions, and through operational history. It’s also the case that every time an employee leaves, the work they created gets lost in a digital abyss.

Changing Work

Now imagine we had a ChatGPT equivalent feeding on an internal corporate dataset. That you could ask this Generative AI to find anything inside the company history, or even better, ask it to generate valuable information based on prompts. It might summarise product launches of the past 10 years finding key correlations of the most successful launches. Or, build next years’ budget document based on previous years insights and deliverables. You may ask it to test our assumptions for next year based on internal and external historical datasets. So much of what we do inside corporations is talking to each other and navigating lost data, an AI like this would change how we work.

Google, Microsoft, Apple

It’s worth noting that this AI integration process has already commenced. I recently wrote about the significance of Microsoft’s investment in ÇhatGPT. They are already integrating ChatGPT-based AI system, Copilot, into Microsoft 365, which is the world’s most ubiquitous corporate software platform. In the first instance, it will impact Word, Excel, PowerPoint, and Outlook by automating tasks like summarizing meeting discussions, creating PowerPoint presentations, drafting emails, analysing long email threads, and generating summaries and graphs on Excel spreadsheets. This means that the ‘training process’ is underway.

And we should expect the competition to come thick and fast, and to go beyond chat bots and into corporate AIs. This will include Google's Bard, Meta's Blenderbot, and Baidu's Ernie, or Wenxin Yiyan.

As investors what we need to be looking for is which AIs are likely to win in each segment. It seems clear that Google and Microsoft are in the best position to unleash internal corporate generative AIs. But it must be said that Apple, which has remained particularly quiet on this (as is their corporate modality – they tend to launch cold) is far better placed for personal AIs.

While Siri isn’t great, it is already deeply integrated into the Apple ecosystem and could surely make a technological stepchange and achieve much higher utility.

It’s time we didn’t just consider which jobs AIs might take away but how they might impact the outcomes of big tech.