The UN is losing credibility as a potential AI regulator

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.

Is Big Tech ruining the ‘AI dream’?

From The New Republic: The current tech industry's approach to AI has deviated from the technology's true promise and early vision. While AI at once seems poised to make lives healthier, wealthier and easier, it’s now also largely being viewed as either as a gimmick (ChatGPT regurgitating training data without having true reasoning capability) or an existential threat. The author argues that the current narrative around AI is Big Tech-led and is widely obscuring its potential real-world benefits.

More details:

• Tech leaders today aren’t doing a good job explaining the benefits AI offers the public (indeed, a recent Salesforce survey found that a shocking 88% of the people who aren’t using AI don’t understand how it could benefit them).

• Smartphones, social media and online platforms broadly have deteriorated via "enshittification" - becoming optimized for ads and shareholders instead of users. This track record contributes to people’s distrust of tech's AI promises.

• The government and Big Tech are allied in the pursuit of economic growth, with AI being co-opted to benefit business interests rather than public interests - similar to past technologies (TV/internet/smartphones).

Takeaways: The tech elite holds a sort of monopoly on narratives about AI's future; that narrative in the news rarely focuses on practical application or reasonable discourse. It would take quite a bit of work for the layman AI user to truly understand what’s going on in the space - good and bad. The public needs a clearer vision of how AI can practically empower and benefit them, not just avoid harming them.

Currently, almost all Big Tech companies offer free courses to ‘learn all about AI’, but these are largely funnels to get users to subscribe to their newest tools and services.

Users can also anticipate AI-enhanced features in apps like Paint and Photos. Furthermore, Bing is getting a notable upgrade with the DALL-E 3 image generator from OpenAI, along with more personalized search results and a revamped AI-powered shopping experience. Microsoft 365's Copilot will be available for enterprise customers from November 1, boasting a broad swath of AI capabilities. Microsoft also showcased a range of new Surface devices optimized for these AI experiences.

Despite this, in “Generative AI’s Act 2”, Sequoia remains optimistic. We’re starting to see generative AI's shift from gimmicky tools and services into focusing on addressing real human problems. One major oversight from Sequoia's initial analysis was the rapid evolution in areas like code generation and video quality, which have advanced far quicker than they anticipated. Amara’s Law - the phenomenon that we tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run - is running its course; Sequoia argues we’re already moving into the long stretch.

• UN deadlocked over regulating AI

• Splunk soars on Cisco takeover in $28 billion AI cloud security deal

• ChatGPT is landing kids in the principal’s office, survey finds

• San Francisco teachers cautiously, optimistically, experiment with AI

• YouTube is going all in on AI

• Researchers in Japan claim AI can translate what chickens are ‘saying’

• Intel’s quantum computing plan to leapfrog rivals

Copilot Chat forms a powerful AI pair with the existing Copilot auto-complete feature. This integration of natural language conversation and code assistance can reduce boilerplate work and might help to eventually establish ‘natural language’ as a new universal programming language. GitHub notes that some of the common use cases for Copilot Chat are real-time guidance to suggest best practices, tips and solutions tailored to the code a developer is currently working on, help with code analysis and fixing security issues - all without having to switch away from the IDE.

• MySQL HeatWave gets generative AI and JavaScript, slew of new features

• (Pinecone) An (opinionated) checklist to choose a vector DB

• ‘Chain of Verification’ reduces hallucinations in LLM’s