Though artificial intelligence is not necessarily new, the field entered a notably new environment when ChatGPT launched in November. Something that had long been under the hood of the internet became visible. And it wasn't long before it became exceptionally popular.
That introduction of an AI chatbot that users are free to engage with seemed to kick off the AI race: OpenAI has ChatGPT; Google has Bard and is working on connecting AI solutions across its Google Suite, something that Microsoft is mirroring. Meta is developing its own AI and Nvidia is surging amid explosive demand for the chips that power these models.
And Microsoft co-founder Bill Gates, speaking at a conference May 22, said that the company that wins the so-called AI race will be the company that develops a personal digital assistant.
Personal.ai is doing just that, and going even further.
DON'T MISS: Human Extinction From AI is Possible, Developers Warn
How It Works
Personal.ai is a personal language model system. Rather than being trained by an enormous data set to produce generic knowledge, Personal.ai is trained by each individual user.
Each user's AI profile is trained regularly on what is essentially a constantly growing memory of that user, including his/her thoughts, beliefs, opinions, etc. The goal is to create an artificial extension of each user.
And that individual data collected over time is used to train that specific user's model; data isn't aggregated. And the ownership of the data remains with the users.
Personal.ai's founder and CEO Suman Kanuganti, inspired by the idea of being able to speak, in some form, to someone who has passed away, is intent on creating more than just a digital assistant.
It can handle tasks, read and write emails and even converse with people to sound like you.
He wants to create digital imprints with the goal of keeping humans in the loop and enhancing human presence and communication.
How It's Different From The Competition
Let's start with ChatGPT.
ChatGPT is a large language model, while Peronal.ai is a small language model. The immediate difference is obvious; ChatGPT is much larger with GPT-3 at around 170 billion parameters, while Personal.ai has only 120 million parameters. (Think of parameters like settings that help a model generate content).
The greater size necessitates a much higher operational cost.
And the creators of ChatGPT, OpenAI, are working to create AGI, a hypothetical that could represent artificial intelligence greater than that of humans. Personal.ai, meanwhile, is working on enhancing human connections, rather than taking them out of the loop.
And, while big-name tech giants like Microsoft and Google are starting to roll out AI assistants across their suite of apps, Kanuganti said that his models are designed to work across platforms and that, with Personal.ai, data sharing is not an issue.
"As a startup, I think we have an advantage of trust. We have an advantage of interoperability," he said. "Meta, Google, Microsoft, they will never share the data with each other."
And because Personal.ai is trained on user data, rather than the internet or aggregate data, issues of misinformation and bias that have been par for the course of AI for a while now, are much less of a problem.
And while Kanuganti is keen on the many ways his small AI model can enhance productivity in all corners of his users' lives, his focus remains on safety and ethical responsibility through data ownership and the democratization of these models.
"Otherwise, all your data is going into Apple, Google, etc. I want to be able to say, 'okay, but now take control.' Put all the goodness that you feel in the digital space in one model," he said. "Personal computing should be personal intelligence. If personal computing you can buy from the store, you can buy personal intelligence for yourself as well."
One Stock We Believe Will Win in The AI Race (It's not Nvidia!)
