Prompt-driven Augmentation/ Regulating Orbital Debris/ How Transformers Seem to Mimic Parts of the Brain/ China’s Factories Accelerate Robotics Push
Massimo Portincaso
Founder & CEO at Arsenale Bioyards, Industrial Romantic and Antidisciplinarian Stoic
Prompt-driven?Augmentation.?I wrote in 2017 about automation and augmentation, in a piece focused on the need for (business) leaders to reshape their narratives. “Most corporate and policy research regarding artificial intelligence and the “future of work” is centered on automation — the replacement of human labor and cognition with machines. Multiple studies report some variation of the same narrative: about half of all jobs in advanced economies may be automated away by 2050, if not earlier.
This stark human-versus-machine dichotomy gives rise to a number of blind spots and neglects important dimensions, such as the spread of complex adaptive systems and the network effects caused by their entanglement. Most importantly, it skips the most promising opportunity space for business and for every sector of society: the human-machine interface.
A narrative of augmentation, instead of automation, invites business leaders, policymakers, researchers, and the labor force to pay much more attention to this middle space. Companies and society need to create a narrative that focuses on the potential of AI to switch the scale of reference for several tasks, often by several orders of magnitude. A good example is personalization. Brands that leverage AI and proprietary data can move from tens or hundreds to hundreds of thousands of customer segments and?see revenue increase by 6% to 10%, two to three times faster than those that don’t harness this potential.
Amazon is a good example of AI as a source of augmentation rather than just automation. The company, one of the heaviest users of AI and of robots (in its fulfillment centers,?the number of robots grew from 1,400 in 2014 to 45,000 in 2016), more than doubled its workforce in the past three years and expects to?hire another 100,000 workers in the coming year?(many of them in fulfillment centers).
The point is that we need a narrative that encourages us to generate more with available (human) resources by leveraging AI and technology, not one that looks at a finite game of optimizing away labor costs wherever they exist.
The augmentation narrative is not limited to products and processes; it also affects professions and management. Just as what it means to be a doctor is going to be reshaped by access to millions of records and machine learning, what it means to be a manager and run an organization will change significantly. The current trend to decentralize decisions will be fundamentally redefined and accelerated as decisions are increasingly supported by AI and data, “augmenting” decision-makers and allowing for new management tools and new organizational structures.“
I have been thinking about the topic since, and it was interesting to see that what I wrote back then fundamentally still applies. But there are some very interesting developments on this front.
One is the development in China, featured below, about the adoption of robots to compensate for the shrinking workforce. There are others though, which warrant more attention, in my view.
The most important one is the step, taken by Google, to start merging Robotics with AI.?Alberto Romero?explores more in-depth in?this post?why this can become a real breakthrough, and he goes on in?another post, to argue the importance of the “prompt” in a world where transformer models have radically changed the landscape of AI and brought a whole new meaning to the notion of language.
So, going back to my article from 2017, I think the most important thought that I would add, if had to re-write it today, is that the augmentation process will be massively influenced by the prompt that we provide to the robot or the AI. This will require us to develop a proper and new way of thinking and expressing the thinking by properly articulating the prompt. This excites me, as it will mean translating creative and articulated thinking into eloquence which will then be turned into augmentation. An exciting prospect, if we manage to master it properly.
Space is getting more “crowded,” increasing the risk of “catastrophic in-orbit collisions.” Two technologies currently deployed to avoid satellite crashes are space situational awareness?(SSA) and space traffic management?(STM). SSA tracks “debris… and active satellites in orbit.” STM predicts “when and where collisions could occur.” Despite these tools, ?Araz Feyzi, Co-founder and CTO at ?Kayhan Space, says, “The reality is that we don’t know precisely where an object is in orbit. It’s in a… cone of uncertainty.”
News items:
A ?recent paper?details how researchers “built a fully independent living structure” by using bacteria to create “highly complex [synthetic] protocells.”
Recent research has shown that the hippocampus - a part of the human brain “critical to memory” - is “basically a transformer?neural net in disguise.” Transformers are the “secret sauce” in LLMs like GPT-3 and BERT and have “since excelled at other tasks such as classifying images — and now, modeling the brain.”
According to cognitive neuroscientist ?James Whittington: “The fact that we know these models of the brain are equivalent to the transformer means that our (AI) models perform much better and are easier to train.”?David Ha, research scientist at Google Brain, says, “We’re not trying to re-create the brain. But can we create a mechanism that can do what the brain does?”
News items:
Could the ?Earth Species Project?- a non-profit “billed as Google Translate for whales” - and the ?Cetacean Translation Initiative?enable us to “communicate with a whale well enough to exchange ideas and experiences by 2026?”
领英推荐
Industrial robot sales to China rose 45% in 2021, accounting for “just under half” of the world’s heavy-duty robot installations, “reinforcing the nation’s status as the No. 1 market for robot manufacturers worldwide.” China’s automation boom is expected to accelerate further, partly due to “the growing difficulty of attracting workers.”?Jay Huang, director of Greater China research at Bernstein, “projects that China will have between 3.2M and 4.2M industrial robots working on production lines by 2030, up from around 1M currently.”
News items:
Disney recently ?filed a patent?that could “allow theme park guests to see 3D AR imagery without glasses.”
Deep tech startups face unique challenges “when considering investors, go-to-market strategy, and communications. Elisheva Marcus, VP Communications at Earlybird Venture Capital, shares 10 PR tips - with real-world examples - for startups with “technology that is inherently complex, and…often difficult to explain.”
Marcus recommends that CEOs become the “face” of the company and cites Isar Aerospace CEO ?Daniel Metzler?as an example. She also advocates for the effectiveness of “lighthouse use cases,” arguing that “the best way to illustrate your technology/product and to make it accessible to the outside world is to communicate the functionalities and USPs by showcasing client stories.”
News items:
Greg Rutkowski: Dragon Cave
AI-generated art has its?first art star, but that doesn’t mean he’s happy about it.?Greg Rutkowski’s work?is so popular and prolific that people have begun using his name as a prompt. Rutkowski doesn’t blame them but says, “For me and many other artists, it’s starting to look like a threat to our careers.”
Clearview AI’s “groundbreaking” facial recognition software has been used extensively by law enforcement to “find criminals” in recent years. Clearview AI’s use of “billions of faces scraped from social media sites… without people’s consent” has made the tool highly?controversial. The company was banned?from selling its database to private companies and fined $9.4M?by the ?UK’s privacy watchdog?earlier in 2022.
Now, Clearview AI “plans to offer access to public defenders” in an effort to “balance the scales of justice,” according to Co-founder and CEO ?Hoan Ton-That. The NY Times highlights the recent case of a man exonerated of a charge using Clearview AI that might have landed him in prison for up to 15 years. But the move is being viewed with skepticism by many.
Jerome Greco, Digital Forensics Supervising Attorney at The Legal Aid Society of NYC, said, “This is mostly being done as a P.R. stunt to try to push back against the negative publicity that Clearview has about its tool and how it’s being used by law enforcement.”
News items:
Currently, 100,000 people in the US are waiting for a kidney transplant - and an AI algorithm decides who’s first to get one. Here’s a deep dive into how “the moral logic of the allocation algorithm” has evolved to be more ethical.