Google Pathways: The Nodes Not Traveled
November 1, 2021
A new post from Jeff Dean, Google Senior Fellow and Senior Vice President, Google Research, posted “Introducing Pathways: A next-generation AI Architecture.” I associate Mr. Dean with Chubby, recipes, and other Google plumbing like spelling variants of the name Britney Spears. Oh, I also recall that he was involved in the non-termination of the Google AI ethics researcher Timnit Gebru and may have provided input to the “find your future elsewhere” approach to Margaret Mitchell. I am not going to walk my French bulldog down memory lane, but these AI ethics people did not get with the Google program and put on a snorkel, a mask, and floppy swim fins. I suppose they wanted to duck walk down a different pathway to a different path to smart software.
The blog post explains that “Google is building a new AI architecture that will handle many tasks at once, learn new tasks quickly, and reflect a better understanding of the world.” Note that the progressive tense is used my Mr. Dean, so the Pathways are still under construction.
The idea, the post explains:
…today’s machine learning models tend to overspecialize at individual tasks when they could excel at many. They rely on one form of input when they could synthesize several. And too often they resort to brute force when deftness and specialization of expertise would do.
Yep, now it is time for weak supervision, packaged AI modules, and orchestration across different tasks. What’s a task? These are not explained in the write up.
Mr. Dean’s write up states:
Pathways will enable a single AI system to generalize across thousands or millions of tasks, to understand different types of data, and to do so with remarkable efficiency…
Mr. Dean emphasizes what seems to me to be an assertion similar to Google’s statements about the company’s quantum supremacy. These fuzzy words sound great and spark buzz, but exactly what’s the yardstick? I don’t know and, of course, neither does Google because this is work which is not completed and explained in Silicon Valley speak.
Mr. Dean makes clear that the recently refurbished staff organization involving DeepMind and the AI Ethics group:
are also sure there are major future challenges we haven’t yet anticipated, and many will demand urgent solutions. So, with great care, and always in line with our AI Principles, we’re crafting the kind of next-generation AI system that can quickly adapt to new needs and solve new problems all around the world as they arise, helping humanity make the most of the future ahead of us.
With Google record profits, a strong belief in the methods crafted in part at SAIL, and the AI professionals who buy into the Google program — the future looks bright.
I like the helping humanity. The “us” seems to reference Googlers.
Sounds good, right? Free email, anyone? Objective search results, anyone? Baked in bias, anyone? I can’t hear you over the industrialized machine learning system. Is there too much bias in those drive belts?
Stephen E Arnold, November 1, 2021
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
