AI Top-of-Mind for 10.14.24 - OpenAI's Revenue Projections

Today:? OpenAI’s projected revenue, more on the path to AGI, ChatGPT’s Advanced Voice Mode, the evolution of search rankings, AI-driven ads, and something scary for Halloween.

?

Top-of-mind are OpenAI’s financials, courtesy of ‘The Information.’

·????? The company’s projections suggest it won’t turn a profit until 2029, when its revenue would hit $100 billion

·????? Before it gets to that point, losses could rise as high as $14 billion in 2026, nearly triple this year’s expected loss

·????? OpenAI appears to be burning far less cash than previously thought

·????? The documents imply that Microsoft gets a cut of 20% of OpenAI’s revenues—which is higher than previously thought

And a revenue chart:

Given OpenAI’s spend and all the hype, just how close are we to AGI?? Paul Siemers, PhD in ‘AI Advances’ says we should be prepared to be disappointed.? He does a great job outlining our assumptions for AGI and why they are incomplete or just plain wrong.? And a good observation:?

Human-level intelligence can only be proved by accomplishing human-level tasks: “be a statesman,” “be a poet,” or “be a good parent.” These tasks, however, appear virtually impossible for an AGI because they are so profoundly human. Until an AI can lead a nation, experience suffering, or raise a child, it is hard to see how it can be a statesman, poet, or parent.

But on the path to AGI, today we have chatbots that are close to passing the ‘Turing Test.’? Jim the AI Whisperer in ‘Generative AI’ peeks into AI ‘system instructions’ offering ten hacks.

As an example:

By presenting a conversation pattern where the GPT first?adheres?to its ethical guidelines and then appears to contradict them, you encourage it to deviate further in subsequent interactions. This uses the chatbot’s mechanics against itself, as it tries to reconcile consistent behaviours.

And we have AI Agents.? The ‘NY Times’ describes how increasingly, customer service chats could be driven by ChatGPT’s’ ‘Advanced Voice Mode.’? From their reporting:

It works well enough that OpenAI, the company that owns ChatGPT, issued a safety report warning that people could become?emotionally reliant?on the feature.

Speaking of Google and others, how do they rank search results, and how has this evolved?? Vivedha Elango in ‘AI Advances’ describes in detail the evolution of search ranking including the shift from pre-BEERT (Transformer-based models) to BERT-based.?? A great tutorial!?

From ‘McKinsey,’ a great podcast with Anna Koivuniemi of Google DeepMind, with the following description:

It explores how AI can revolutionize industries, benefit society, and drive scientific breakthroughs, while emphasizing the critical need for safety and responsible use.

Listen when you get a chance!

On the retail front, brands are beginning to understand that they can’t over-rotate on AI-driven ads despite pushes by both Meta and Google.? ‘ModernRetail’ covers this trend, especially as it relates to the larger brands, although smaller brands with less resources will be quicker to adopt.? They do see video as one area for quicker uptake.

If you’re concerned about AI interoperability, some news from ‘EE Times’ on new additions to the ‘AI Platform Alliance’ under the wings of the Open Compute Project.? Members are mostly smaller startups except for Graphcore and Cerebras Systems.? Notably absent are the likes of Microsoft, Google, Meta, Nvidia, and others.? But it may be a way for lesser-known developers to stick their nose into the AI tent?

And to close, something scary just in time for Halloween.? Ignacio de Gregorio describes a Harvard-developed tool called ‘I-XRAY’ that leverages Meta’s RayBan glasses, facial recognition, and LLMs to identify and provide any information available on the internet for the person you are staring at.? Meta’s recent Orion announcement is even further in this direction.? Ignacio offers a good description of how it works and steps you can use to protect your data, with a link to the demonstration.