'Centaurs and Cyborgs' - two emerging types of AI users

Plus: Combining natural language with autonomous vehicles

Author

The Dispatch
September 19, 2023

Welcome to The Dispatch! We are the newsletter that keeps you informed about AI. Each weekday, we scour the web to aggregate the many stories related to artificial intelligence; we pass along the news, useful resources, tools or services, technical analysis and exciting developments in open source. Even if you aren’t an engineer, we’ll keep you in touch with what’s going on under the hood in AI.

Good morning. Today in AI:

  • Columbia law professor Tim Wu writes in The New York Times about the stakes of Google’s antitrust trial

  • ‘Centaurs and Cyborgs’ - new Harvard-led research likens those who successfully use AI to mythical creatures

  • C3ai CEO Thomas Riebel praises California Gov. Gavin Newsom’s exploratory AI legislation

  • Financial Times looks at the current state of AI in banking

  • AI startup Wayve is integrating natural language with autonomous vehicles

  • Palantir CTO Shyam Sankar shows how Palantir is using LLM’s in their AI platform, Elon Musk gets interviewed at the All-in Summit, trending tools & more

A recent Google advertisement in Shanghai Image: Shutterstock

From The New York Times (Opinion): The Google antitrust trial (linked: NPR’s coverage) has captured attention for probing the search giant's past business practices, notably its deals with Samsung and Apple to secure a dominant position in online search. However, the stakes extend far beyond past actions and encroach upon the future of tech competition itself. The verdict in this case will set precedents affecting not just the world of online search but also commercialized AI - and possibly technology we haven’t even developed yet.

More Details:

  • Setting future norms: Loosening the grip of a controlling monopolist may not always solve the problem at hand. But it can open up closed markets, shake up the industry and spark innovation in unexpected areas.

  • AI and Google: Google is not merely defending its past but also safeguarding its extensive investments in AI. Google has spent billions on AI research and product development, with the objective of maintaining its competitive edge.

  • Prosecution's case: The prosecution argues that Google leveraged its financial might to suppress competition, specifically paying Samsung and Apple billions to make Google the default search option on their phones. Apple also agreed to stay out of the search business.

Takeaways: The Google antitrust trial could be a watershed moment, much like past landmark cases against AT&T and IBM, which reshaped industries in unintended ways. These previous cases opened up closed markets and spurred innovation. Google's case also bears an echo of irony: once a beneficiary of antitrust actions against Microsoft in the early 2000’s, Google now finds itself in the crosshairs, facing scrutiny for tactics eerily similar to those it once capitalized on.

A research project led by Harvard social scientists and Boston Consulting Group studied the impact of AI on the future of professional work. The findings, discussed here through the eyes of one of the researchers, overwhelmingly suggest that AI significantly boosts productivity, task completion rate, and - perhaps surprisingly - 40% higher quality of output in a consulting setting.

The study introduces the concept of the "Jagged Frontier," illustrating that AI's capabilities are uneven and can lead to hard-to-notice mistakes. However, workers who adopted one of two approaches (either as "Centaurs," who strategically divide labor between human and AI, or as "Cyborgs," who integrate human and machine tasks) were found to best navigate the complexities and inconsistencies of AI capabilities. The study indicates that AI is already a game-changing technology in professional settings and emphasizes the need for ethical and strategic adoption of AI in the workplace.

In the banking sector, the use of AI has not quite led to the dramatic workforce reductions that some leaders predicted a few years ago. Despite Deutsche Bank's former CEO John Cryan forecasting in 2017 that robots could replace about half of the bank's 98,000 employees, the actual headcount reduction has been around 10%, largely due to routine cost-cutting rather than AI.

Moreover, employment in sectors like the New York financial securities industry has actually risen almost 8% between 2017 and 2022. Industry executives still view AI as a core driver for efficiency and customer engagement; but they’re focused more on incremental gains in productivity rather than eliminating positions. From HSBC’s chief operating officer John Hinshaw:

“This [embracing of AI] is not about job elimination, it’s about . . . doing more with the same number of people and having a more fun place to work in a lot of ways.”

From our sponsors:

Scammers and Spammers Hate Aura - and That's a Good Thing for You

Are you tired of unwanted emails, texts, and calls from scammers and spammers? Aura uses cutting-edge AI technology to scan the internet to identify where your personal information is exposed and being sold. With Aura, you can take control of your online privacy and secure your personal information. Reduce robocalls and keep your information safe from identity thieves. Try Aura free for 2 weeks and see if your information has been compromised.

-

AI Startup Wayve is introducing a new frontier in autonomous driving technology by building on Vision-Language-Action Models (VLAMs). VLAMs in AV’s aim to blend three types of information: visual data, driving data, and natural language, to train foundation driving models that not only perceive and plan but also comprehend and reason.

These models open up the exciting possibility of interacting with driving systems through dialogue. Users could ask autonomous vehicles about their actions and rationale, and provide natural language instructions. For example: "move into the right lane when it is clear," or "drive more slowly, there are pedestrians ahead."

The use of natural language in the training phase may enable faster training and better generalization to new driving environments, enhancing the capabilities of these models in real-world applications. Wayve is still addressing limitations with scaling the platform and issues like LLM ‘hallucinations’.

Trending AI Tools & Services:

  • Coverposts: enter your article URL, and Coverposts will output perfect fit social media content with images and illustrations

  • Lingtual: build AI tools and chatbots that connect to everything you use (calendar, email, company API's, docs, DB's, and more) - without code

  • Screenwriter: build, execute, and scale UI tests - 30x faster, zero code, and flake free

  • Brainglue: unleash the power of AI with prompt chains

  • Fireworks AI: the world's fastest LLM inference platform. Use a state-of-the-art, open-source model or fine-tune and deploy your own at no additional cost

Guides/useful/lists:

Social media/video/podcast:

  • How to confront the 'AI dilemma' - POLITICO Tech [Podcast]

  • (Discussion) Google changes its stance on AI generated content [Reddit]

  • More experiments with HeyGen’s incredible video language translations [X]

  • AIP - How Palantir is using LLM’s [YouTube]

  • All-In Summit: Elon Musk on Ukraine, X, the creator economy, China, AI, & more [YouTube]

Did you know? 

Jack Dongarra, a Turing award laureate and co-founder of TOP500, has cautioned that the United States may be falling behind China in the arena of exascale supercomputing. According to reports, China's third exascale supercomputer has recently come online, despite initial thoughts that the project had been halted due to U.S. sanctions. In contrast, the U.S. has only two operational exascale machines, Frontier and Aurora. China does not participate in the official TOP500 list - likely due to fears of attracting more U.S. sanctions - and Dongarra suggests that China's capabilities are likely underrepresented in such rankings.

The situation casts some doubt on whether current sanctions are fully hindering China's technological advancements as intended. This third Chinese exascale machine, reportedly developed by China-based Sugon, resumed operations even after the company was placed on a blacklist in 2019 - which resulted in losing access to key processors.

❝

Maybe having the No.1 computer would make news and put China under the spotlight. It can cause the US to take actions against China that would further restrict technologies from flowing into China.

Jack Dongarra, supercomputer guru, September 2023