Anticipating Risks & Opportunities of Generative AI: Prompt Engineer, Larsen Effect, WestWorld Syndrome, Copyright Infringement…

These past months, I participated in panels, podcasts, and keynotes on #GenerativeAI, focusing on anticipating risks and opportunities for individuals and organizations to start their Gen AI ride with confidence and care. I would love your thoughts on those:

1- Regarding #FutureOfWork, a lot of press announces the end of creative, intellectual, and white-collar jobs, with the rise of the #PromptEngineer as the new dream job — well paid will be those who can whisper to the machines! But more than thinking about #Prompting as a new job, we might want to consider prompting as the new must-have skill. #chatGPT and likes do create disruptive innovation similarly to search engines or emailing did a couple of decades ago. Noone’s only job is to conduct Google search or write emails as a lone task all-day long. Those have become everyday skills integrated within our work processes. Given the level of complexity of certain prompts and the impressive results we can extract from the latest models, we can still wonder if prompting for generative content will remain a mundane skill or eventually get automatized itself — with prompt shortlisting, parameters, filters. #PromptMarketplaces are monetizing prompt at around 4 dollars.

2 — In terms of #FutureOfCreativity: We anticipate the #LarsenEffect (feedback loop) or the amplification of noise until it becomes painful. What happens when new content is generated out of generated content itself based on generated content — what percentage of human creativity remains present in the output? Researchers recently found irreversible defects in the resulting models, causing #ModelCollapse. Also, imagine an angry customer asking chatGPT to create a message for customer support… which is addressed by customer agents themselves generating automatized responses. In this chatty dialog among AIs, will we still be able listen to human voices in the noise? And most importantly — what will be the cost of those extensive chats on the #Environment?

3 — Future of Content Consumption: As a consequence, we might end up living in a world of generated assets (#WestWorldSyndrome) — news, advertising, journalist voices, influencers, actors. Godfather of AI Geoffrey Hinton quit his job at Google, expressing concern that the internet will be flooded with false photos, videos and text, and the average person will “not be able to know what is true anymore.” As Generative AI continue to get popular as an #ActionEngine, our sense of reality might be truly at stake.

4 — #Ethics and #Responsibility: Ethical implications are immense, and we must anticipate them early to prevent worst-case scenarios from happening. We already hear about voice #IdentityTheft to access bank accounts, children manipulated voices asking parents for ransom, big music labels taking action to remove generated content using their artist’ performance, voice, or appearance, while Japan considers that AI model training doesn’t violate #Copyright.