December 2024
AI Part 9
OpenAI to Debut Limited AGI in ‘25

The definition of artificial general intelligence (AGI) may be subject to debate, but the main thing that developers and AI companies can agree on is that it is far superior in terms of speed and precision than current AI models. Is it really ‘thinking like a human’? Not exactly. But if it does what OpenAI predicts it will, it will still blow our minds when it becomes available next year.

OpenAI has been testing its newest effort, a blend of its Orion and Strawberry models, but with far more safety protocols and guardrails. According to Sam Altman, its agent capabilities will be central to its advancement and expand its context window. While it won’t be able to reason like a human — yet — the company and others expect to build in recursive and lateral thinking over the coming couple of years. Right now they’re refining core reasoning through inference optimization and synthetic data generation.

The target markets for this low-level AGI will be big corporations for the most part, but in a few years costs will decline so much that anyone will have access to systems so efficient they’ll conduct rapid problem-solving far better and quicker than any human could. That means a lot more people will be forced to learn how to manage AIs rather than people — because they won’t need to.

Along with OpenAI, Microsoft AI is leading the way with training and capabilities enabling “infinite memory” for users, says CEO Mustafa Suleyman. “Infini-Attention” means it will remember all the steps used by people creating agents and all its inputs and corrections, and build on that in its execution by agents, combining the “short-term and long-term memory” of the project.

Once an agent is taught, it will be the equivalent of a personalized computer system that Microsoft AI anticipates could save a user as much as 20 hours of work a week, for instance. In as few as five years it will allow us to automate most of the work that we do, potentially impacting the efficiency of law work, accounting, technology development and so much more. Right now the error rate is too high to make this practical.

Eric Schmidt, a CEO of Google and now founder of the Special Competitive Studies Project (SCSP), a nonpartisan US think tank and private foundation focused on technology and security, says “infinite context” will be the game-changer in AGI.

Imagine a computer that remembers every little thing you tell it. Here's a link to a video that explains it:

https://youtu.be/g1JSZwwtEak.

Tool for podcast research

Podcasters spend much of their time researching information and writing scripts or integrating the information during recorded conversations among two or more people that they later edit. Imagine using an AI tool that does the research for you, and with a few prompts, integrates it into a script. a story in MIT’s Technology Review magazine asserts that it’s available now with the tool Audio Overview by Google.

Not only can the software make it far easier to put together a podcast, it has the capability of using humanlike synthetic voices to create it on the fly. Some people are already using it to summarize books and listen to their own personal podcast that explains the book’s content far faster than they could learn it on their own.

I’m not saying that this appeals to everyone. A radio station in Poland decided to fire its announcers and use AI announcers instead, but the public uproar that ensued made them decide to return to their former system. This kind of tool can be helpful if used properly, saving podcasters hours of work. They can take or leave the writing of the podcast tool, and craft the podcast content according to their own styles.

Once again, this is an AI program meant to make lives easier and save time.

Writing, tech, graphics jobs take AI hits

A study by researchers at the Harvard Business School, German Institute for Economic Research and Imperial College London Business School found the first solid evidence that AI is replacing freelance jobs.

In the joint analysis of more than 1.3 million job posts from 2021 to 2023, the researchers found a 21% decline in demand for jobs like writing and software development just eight months after the release of ChatGPT. In addition, they discovered a 17% decline in demand for graphic design and 3D modeling work.

Not only are such jobs disappearing, the pay rates for freelance jobs are dropping too, as the competition increases. The researchers say worker value lies in creativity, critical thinking, and emotional intelligence.

The good news is that people who learn AI-system management will be in demand. Consultants who currently have those skills are already being snapped up.

Journalist Toni Denis is a frequent contributor.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.