AI Revs the Existential Risk Engine

This week's edition of Tomorrow Bytes provides an in-depth look at the rapid evolution of AI and its growing influence. Key developments covered highlight generative AI's expansion into new frontiers like personalized assistants and app prototyping while also examining emerging challenges like unrealistic beauty standards and existential risks. We gain perspective on AI's business promise and limitations, as well as the complex dynamics around ethical governance. The newsletter distills critical insights on how AI breakthroughs may shape our future across industries, research disciplines, and policies. It synthesizes diverse signals to reveal the accelerating pace of AI progress and the nuanced considerations needed to steer it responsibly.

As AI becomes even more interwoven into our daily lives, Tomorrow Bytes is here to shed light on its transformative journey. Let’s get into it…

📌 Byte-sized Breakdowns

• Google's updates to Bard will revolutionize digital assistants by having more natural conversations.

• Microsoft has crossed a major milestone with over 1 million paying users now leveraging its code-generating AI tool GitHub Copilot.

• Amazon is bringing more natural conversational AI to kids through a new feature called Explore with Alexa for open-ended learning.

• Google Photos will give users more personalized control over the AI-generated highlight videos it creates, responding to complaints.

• Qualcomm integrated on-device AI capabilities directly into its latest chips, allowing AI experiences without connectivity for phones and PCs.

• The FCC is proactively investigating how advances in AI may impact the ongoing fight against illegal robocalls and robotexts.

• San Francisco is experiencing renewed energy in AI research and development, attracting talent and investments.

• A biotech company, Owkin, developed an AI platform to help doctors securely share research insights, accelerating discoveries and cures.

• Midjourney launched a new beta website to showcase AI-generated images and make it easier for users to discover art prompts.

• Boston Dynamics introduced conversational AI capabilities for its robots, allowing natural dialogs to direct their actions.

• Shutterstock integrated AI image generation into its library, offering over 700 million creative images powered by DALL-E.

• Google's leaked Gemini project suggests the development of a multimodal AI tool that can generate app prototypes based on user prompts, potentially revolutionizing the design process for mobile applications.

🔦 Spotlight Signals

US halts AI chip exports to China: The US government has ordered Nvidia to stop exporting two top AI chips to China and Russia, escalating tensions between the superpowers over technology. [Dive In]

White House to unveil AI order: The Biden administration plans to issue an executive order next week laying out guidelines for how to regulate artificial intelligence technology and promote innovation. [Dive In]

Frontier Model Forum launched: A new forum called Frontier Model aims to facilitate communication between institutions developing the most advanced AI systems. [Dive In]

AI has rescued Mark Zuckerberg from a Metaverse-size hole: Meta's heavy investments in AI research are starting to pay off, with generative AI offering new opportunities after overspending on Metaverse dreams. [Dive In]

Google and Apple clash over search and spotlight: Google is objecting to Apple giving its Spotlight search priority, escalating tensions between the tech giants as AI transforms web search. [Dive In]

The Beatles: ‘final’ song Now and Then to be released on 11/2: An unfinished Beatles track has been completed using AI to mimic the band's style, demonstrating music generation tech's rapid advance. [Dive In]

Google Reports 7x Growth in AI Projects Built on Its Platforms: Google Cloud saw major expansion in AI workloads as organizations build custom models using its AI services and tools. [Dive In]

Outset is using GPT-4 to make user surveys better: Startup Outset is leveraging Anthropic's new GPT-4 model to generate survey questions that are more engaging and insightful. [Dive In]

Google Maps is getting new AI-powered search: Google Maps is gaining AI-enhanced discovery features, more detailed navigation maps, and expanded availability of its Immersive view. [Dive In]

Bill Gates feels Generative AI is at its plateau: Bill Gates believes the rapid progress in generative AI models like GPT-3 is slowing, with future iterations like GPT-5 offering diminishing returns. [Dive In]

Prime Minister's speech on AI: The UK Prime Minister gave a major speech on AI regulation ahead of next week’s AI Summit, arguing new rules are needed to capture the technology's benefits while minimizing risks. [Dive In]

Apple’s job listings suggest it plans to infuse AI in multiple products: Apple is ramping up AI talent across many teams, signaling ambitions to integrate advanced AI capabilities into its hardware and software. [Dive In]

💼 Business Bytes

OpenAI Faces Growing Competition in AI Market

OpenAI is seeing pricing pressure on its AI offerings as corporate customers explore cheaper alternatives, per The Information. High costs of its API access have deterred some clients amid economic uncertainty. This highlights increasing competition in generative AI as rivals seek to undercut OpenAI. [Dive In]

AI's Near-Term Business Impact May Be Gradual

While generative AI has impressive capabilities, the FT argues practical limitations may curb its capacity for near-term business transformation. Issues like computational costs and real-world complexities will take time to address. So its disruptive impact in the next 5-10 years may be more gradual. [Dive In]

Ad Market Rebound Underway After 2022 Declines

Strong quarterly earnings reports from Google and Snap indicate a rebound in online advertising spending after declines in 2022 amid economic uncertainty. Analysts attribute the recovery to increased ad spending in sectors like travel that scaled back last year. The upbeat results signal renewed optimism about advertising growth heading into 2023. [Dive In]

Microsoft CEO Discusses AI Ethics, Activision, Economic Outlook

In a recent interview, Microsoft CEO Satya Nadella said closing the pending $69 billion Activision Blizzard acquisition remains a priority and expressed confidence it will go through despite FTC objections. He also discussed efforts to ensure responsible development of AI technology. Overall, Nadella remains upbeat on Microsoft's outlook despite broader economic uncertainty. [Dive In]

☕️ Personal Productivity

Time Tracking Made Easy

The free time tracking app Reclaim AI uses advanced automation to track your meetings, tasks, and habits with no manual entry needed. By organizing your time data, it provides insights to boost productivity. [Dive In]

AI That Mimics Your Writing Style

Grammarly's new Generative AI feature studies your writing samples to learn your style - word choice, sentence structure, tone - then generates new text that sounds just like you wrote it. This personalization brings AI writing assistants closer to replicating an individual's unique voice. [Dive In]

Chatbots Igniting New Ways to Explore Literature

Rather than passively reading books, AI chatbots let you have active dialogues about them - asking questions, receiving summaries, gaining insights. This innovative application allows deeper engagement with literature without reading it directly. [Dive In]

Vivid Glimpses of Potential Futures Through AI Art

DALL-E 3 generates stunningly realistic images from text prompts, essentially bringing imaginations to life visually. By conceptualizing future scenarios, it reveals vivid depictions of what could be. This provides thought-provoking visions of potential futures. [Dive In]

🎮 Platform Plays

Lenovo Commits to Enterprise AI With Nvidia

Lenovo outlined their strategy to deliver end-to-end artificial intelligence solutions for enterprises, including a partnership with Nvidia for accelerated computing. Lenovo also announced new ThinkSystem severs optimized for AI workloads, with features like increased memory bandwidth. The announcements demonstrate Lenovo's focus on providing an integrated AI platform. [Dive In]

Qualcomm Enables On-Device AI With New Chipsets

Qualcomm unveiled its latest Snapdragon mobile chipsets and ACx PC processors with integrated AI acceleration, enabling more advanced AI experiences directly on devices. This allows for faster inference and protects privacy by keeping data processing locally. New AI capabilities supported include semantic segmentation, super resolution upscaling, and multi-frame noise reduction. [Dive In]

Amazon Ads Get An AI Visual Boost To Improve Results

Amazon launched AI-powered image generation capabilities for sponsored ads, helping sellers quickly create high-quality visuals. The new feature uses text prompts to generate unique product images. This aims to improve customer experience by showing more realistic and appealing visuals, ultimately driving increased click-through and conversion rates. [Dive In]

Google Fights Misinformation With New AI Image Verification Tools

Google introduced new algorithms that can detect manipulated images and help determine authenticity. The tools will be available in Google Image Search and Chrome. Google is providing these capabilities as misinformation through doctored photos and deepfakes continues to spread widely online. The move aims to give users more power to fact check images. [Dive In]

🤖 Model Marvels

AI Reaches New Heights in Language Skills

The latest breakthrough in natural language processing allows AI to generalize knowledge and apply it to new contexts. This level of language understanding brings AI one step closer to human-level intelligence. Research like this from AI pioneer Geoff Hinton pushes the boundaries of what machines can comprehend. [Dive In]

Jina's 8K Embeddings Give OpenAI a Run

Jina AI open-sourced an 8,000-dimensional text embedding model to rival OpenAI's CLIP. Having more open models like this available levels the playing field for startups to build highly capable AI. Jina's model shows the power of the open-source community to stand alongside Big Tech's offerings. [Dive In]

Generative AI Makes Robots Smarter and More Capable

Advances in generative AI are making robots better at planning, problem-solving, and adapting to novel situations. Models like Anthropic's Claude can enable robots to understand natural language instructions and execute complex tasks. This technology makes robots more ready for widespread adoption. [Dive In]

AI Experts Call for Responsible Innovation

Pioneers Yoshua Bengio and Geoffrey Hinton recommend tech companies devote a third of AI resources to safety. Managing risks from advanced AI is crucial as the technology grows more powerful. Thought leaders driving innovation have a responsibility to steer progress down an ethical path. [Dive In]

🎓 Research Revelations

Teaching AI New Tricks More Efficiently

Researchers at Anthropic introduced AgentTuning, a new technique that allows large language models like ChatGPT to rapidly acquire new skills and knowledge like a human, without requiring extensive fine-tuning. AgentTuning reduces the computational cost of training by 10-100x. This enables models to learn faster and opens up many more applications like personalized assistants, tutors, and domain experts. [Dive In]

Smarter Techniques to Evaluate and Improve AI

Researchers from Google Brain propose Branch-Solve-Merge, a new method that uses tree search algorithms to significantly improve evaluation and generation from large language models like GPT-3. By branching and merging text, it reduces errors, bias, and contradictions while increasing coherence, contextual awareness, and factual accuracy. This technique could help make AI systems more robust, safe, and aligned with human values. [Dive In]

Examining the Existential Dangers of Advanced AI

As their AI systems like GPT-3 and DALL-E become more powerful, OpenAI has formed an internal team to study potential risks like an AI becoming misaligned with human values and causing harm. This proactive approach to AI safety research is prudent and needed, as uncontrolled, superintelligent AI could pose an existential threat in the future. More dialogue and planning around AI ethics is important as the technology advances. [Dive In]

Addressing AI's Proclivity for Unrealistic Beauty Standards

Analysis shows AI image generators like DALL-E 2 and Stable Diffusion often depict unrealistic, eurocentric beauty standards favoring young, thin, white, able-bodied subjects - likely reflecting bias in the training data. Researchers recommend diversifying the data and carefully tuning the algorithms to encourage more varied outputs. Fixing this bias is important for inclusive, ethical AI design. [Dive In]

🚧 Responsible Reflections

AI's Looming Reckoning

Google AI head Demis Hassabis issued a stark warning that artificial intelligence could pose an existential threat to humanity if misused, stating AI is "the most powerful technology that humanity will ever invent." His grave concerns underscore the urgent need for thoughtful governance and oversight to ensure this transformative technology benefits society. Responsible AI development will require coordination between the public and private sectors. [Dive In]

Guiding AI Like Nuclear Power

A new proposal suggests modeling AI oversight after frameworks used to govern nuclear energy and other high-risk industries, hinting at AI's potential societal impact. However, unlike mature technologies, AI changes rapidly, so oversight will require flexibility and agility to adapt to innovations, not rigid regulations. Striking the right balance will be key. [Dive In]

Self-Governance's Limits for AI

The White House secured voluntary AI ethics and impact assessment commitments from tech giants like Microsoft and Amazon. However, history shows binding public-private governance is often needed over time as economic incentives grow. While self-regulation is a start, formal oversight will likely be essential as AI advances. [Dive In]

Patience Needed for AI Governance

Politicians face pressure to quickly regulate AI before harm occurs, but regulatory haste risks stifling innovation essential for proper oversight. The Economist argues more deliberate, flexible policymaking, though slower, will better serve society long-term than rushed regulations. [Dive In]