We replaced our CEO with AI for a day
Last Friday, we officially closed our first quarter with our traditional Quarterly meeting. It is the time and place to celebrate our achievements and solidify our vision and strategy for the quarters to come.
Of course, Generative AI took centre stage. The ML team showcased some of the new use cases that we added to our portfolio, our practices shared how they use it in their day-to-day way of working, our CTO clarified the legal framework we are navigating and, finally, our CEO Jeroen, called in to present the way forward…
Except for the fact that Jeroen didn’t prepare any speech and was flabbergasted when his voice blasted to the speakers to proclaim the way forward for Generative AI at In The Pocket. In reality, the ML team had copied Jeroen’s voice and used it to replace our CEO at the quarterly meeting.
Have a listen yourself:
How about that? Just to keep you informed, we went through the evolution of the experiment, starting from 30 minutes and 18 hours of training, before eventually settling on the final result after 24 hours of training.
After some time, suspicion started to grow. Although Jeroen LemAIre’s tone of voice was correct, the old guard recognised that the intonation and way with words weren’t Jeroens finest. Let’s see if you are able to detect which recording is fake.
Are you able to detect the difference in these shorter snippets?
If you detected some more humanity in the first snippet, you can still distinguish the real stuff from AI. Now of course, we don’t think that this will replace Jeroen anytime soon. After all, he is more than just a voice. But it made us think about how to use and misuse Generative AI technologies like Large Language Models & Diffusion models. This fun little experiment mainly learned us two things:
1. Generative AI technology is here to stay and ready to be used.
This gimmick clearly showed that the state of AI today is powerful to be used in many real-life settings. And we can only expect this technology to improve. Today we used Jeroen’s voice to speak in English, but as In The Pocket grows and spreads to other countries, we might want to translate Jeroen’s voice and message in more languages. The other way around, voice-to-text, will help us become more inclusive and lower the barrier to collaborating easily.
2. The time is right to build AI systems with purpose.
Today, there are still a lot of unanswered questions regarding the power of Generative AI in general. What about IP or consent regarding the input of these models? Who owns a trained model and its output? Are we enabling or blocking the creative process? What are the ethics of creating and using this content? There is no single right answer to these questions.
These matters are both complex, nuanced and often context-dependent. That doesn’t mean that we should stop searching for answers, but we should always be purposeful. When designing solutions and developing products, we should take into account how we use Generative AI and why they benefit the goal of these products.
There will always be secondary effects that we can’t anticipate, but at least we should make sure our primary goal is meaningful and aims to create a positive net impact. For instance, for our voice demo we had clear consent, we knew what we wanted to do and why.
Below you find some more examples of how we use Generative AI purposefully in our organization and how we consider both risks and rewards.
And, for now, the real Jeroen will probably remain our CEO.
