The proof is in the pilot: Why AI process experiments can’t wait, according to futurist Steve Brown

https://delivery-p141552-e1488202.adobeaemcloud.com/adobe/assets/urn:aaid:aem:4663d272-c786-4f23-9499-46c71bd5cae9/as/240905_Face-Value_Steve-Brown_AI-Event-02.png

“Are you fascinated by AI and maybe a bit tired of ChatGPT?”, asks Steve Brown, noting the show of hands raised in the audience at One World Trade Center.

The futurist is laying out the case that AI’s enterprise honeymoon is giving way to a proof-of-concept era where trust is essential and use case selection is crucial. Brown, futurist, entrepreneur, and Google DeepMind alum, reassures us that AI collaborators will help us do our jobs more effectively. His advice for companies seeking to leverage AI to outcompete their rivals: Start piloting now.

”If you’re not investing in AI, you’re the one that’s getting left behind,” cautions Brown.In a talk for our Face Value series, Brown draws on a quarter-century of experience in AI and autonomous agents, including tenures as Chief Evangelist and a futurist working in Intel Labs, plus a more recent stint at DeepMind, Google's AI research lab.

Stay “robot-proof”, stay ahead of competitors

“Remember, today AI is as bad as it's going to be. It’s only going to get more and more capable over the next few years,” says Brown. “If you’re not already running AI pilots and experiments then you need to start doing that.”

One company bucking that trend, he says, is JPMorgan, which has launched a number of generative AI deployments. Others should follow suit, he advises, and that urgency should extend across all 3 dominant AI types now in use: discriminative, generative, and agentic. Let’s take a look at the three models.

Why the 3 types of AI matter: Discriminative, Generative, and Agentic

  • Discriminative. “This is classic machine learning,” says Brown. These supervised AI models typically distinguish among different classes or categories within a labeled dataset. Think detecting objects, recognizing faces, distinguishing between words, splitting Netflix viewers into taste groups, and nudging Amazon customers with ‘you might like...’
  • Generative. Models that create realistic content – such as text, images, code, or audio – by making use of multimodal large language models and image generators through a technique called diffusion. Think how Dall-E generates realistic images and art from natural language prompts, or how ChatGPT can write, debug, and explain code in multiple programming languages.
  • Agentic. One of the latest advances in the field, Agentic AI pushes beyond just classifying data or generating content to independently perform tasks and make decisions through goal-oriented behavior.