The AI Power Paradox

Also: There’s Too Much Money Going to AI Doomers

Author

AI KATANA
August 17, 2023

Hi!

Today's insights shed light on the balancing act of AI governance in a competitive global landscape, the disproportionate funding for AI existential risks versus current challenges, and cutting-edge AI implementations ranging from the tennis courts of the US Open to WhatsApp's sticker feature. We also delve into tool updates, including Google's enriched search experience, and the latest venture capital moves. Stay informed on the pivotal role AI plays across sectors, its promises, and its predicaments.

Sliced:

  • 💪🏽 The AI Power Paradox

  • 📛 There’s Too Much Money Going to AI Doomers

  • 🎾 IBM and the USTA Add Generative AI Commentary and AI Draw Analysis to the 2023 US Open Digital Platforms

  • 😉 WhatsApp starts testing AI-generated stickers

The rapid advancement of AI poses immense opportunities as well as risks, and current governance frameworks are ill-equipped to deal with it. AI differs from past technologies in its speed, power, ubiquity, dual-use potential and ability to undermine state sovereignty. An uncontrolled AI race between the US and China seems likely. To govern AI, the world needs tailored solutions based on principles of precaution, agility, inclusiveness, impermeability and targeting. This could include a scientific body to advise on AI risks, arms control-like approaches to advanced systems, and a financial stability-inspired entity to respond to crises. Getting governance right will require rethinking assumptions about technology and geopolitics. If done well, it can promote AI’s benefits and prevent its harms. The alternative, uncontained AI, poses unacceptable risks. There is no time to waste in building effective cooperative governance.

In the midst of the AI revolution, there's a stark disparity in funding between organizations addressing present-day AI issues and those focused on potential future existential risks. Contemporary safety initiatives, like the European Digital Rights Initiative and AI Now Institute, operate on comparatively modest budgets. In contrast, entities exploring existential threats of AI, such as the Future of Life Institute and the Center for AI Safety, have received millions from prominent donors. Open Philanthropy stands out, pouring nearly half a billion dollars into rogue AI research. This imbalance, in part, may be due to ideological and commercial reasons; more powerful AI models are appealing for investments. Startups, such as OpenAI and Anthropic, which initially committed to AI safety, have later transitioned to commercial enterprises with significant investments. Meanwhile, immediate issues, like AI's tendency to sexualize women, particularly women of color, are often sidelined. The current funding dynamics suggest a need to re-evaluate priorities, ensuring that present-day concerns receive ample attention and resources.

****IBM, in partnership with the United States Tennis Association (USTA), has introduced new digital features for the 2023 US Open. These enhancements include AI-generated commentary and AI Draw Analysis. Using IBM's watsonx AI and data platform, the commentary feature will automatically generate detailed audio narration and captions for US Open highlights from every singles match across all seventeen courts. This was achieved by training AI models specifically on the language of tennis. In addition, the IBM watsonx powered AI Draw Analysis provides an innovative tennis statistic, analyzing and projecting player advantages in the singles draw, updated daily as the tournament unfolds. Both features aim to offer fans deeper insights and a more engaging experience. IBM has served as USTA's technology partner for over 30 years, constantly innovating to elevate the digital experience for millions of fans globally.

Meta-owned messaging platform, WhatsApp, is testing an AI-generated sticker feature in its latest Android beta version. This feature enables users to create stickers based on textual descriptions. The underlying AI model powering this capability has not been specifically identified, but it appears to function similarly to models like Midjoureny or OpenAI’s DALL-E, which generate images from text. Given prior concerns with misleading content spread on WhatsApp, the platform has a reporting mechanism for inappropriate stickers, but it remains unclear what specific safeguards are integrated into the AI model itself. WABetaInfo suggests that these AI-generated stickers will be easily identifiable, possibly using a watermark or similar indicator, akin to Microsoft's clear labeling on images created via OpenAI's DALL-E for Bing Chat. Instagram is also said to be developing labels for AI-generated content.

🛠️ AI tools updates

Google has updated its Search Generative Experience (SGE), an AI-enhanced conversational search feature, to further assist users in understanding and navigating online information. The new tools offer definitions for unknown terms, coding assistance across languages, and the ability to tap into SGE while browsing. Users can now hover over specific words to view definitions, images, and diagrams, a functionality available for various subjects. This mirrors a similar feature on Bing Chat. For those seeking coding aid, SGE will now color-code code snippets for clearer comprehension. Another significant development is the experimental "SGE while browsing" available on Google's mobile apps and soon on desktop. This feature assists users in navigating long-form content, presenting AI-generated key points and Q&A lists. Although similar to existing content highlighting features, this iteration is AI-driven. While Bing Chat offers a comparable function, Google’s version won't summarize paywalled content. Beyond Google and Bing, the news app Artifact, created by Instagram's founders, also incorporated an AI summary feature in 2023. The recent enhancements join other SGE features like video and image displays related to search queries, available through the Google app and Chrome.

💵 Venture Capital updates

Voiceflow, a Toronto-based platform specializing in the creation of conversational AI experiences, has secured $15 million in a funding round spearheaded by OpenView, raising the startup's valuation to $105 million post-funding. Established in 2019 by Braden Ream, Tyler Han, Michael Hood, and Andrew Lawrence, Voiceflow offers a comprehensive platform that empowers teams to design, trial, and introduce AI agents, supporting integration with various natural language understanding platforms. Despite competition from major players like Google's Dialogflow and Microsoft Bot Framework, Voiceflow boasts a robust client roster, with notable names like Amazon, JP Morgan, and The Home Depot, accumulating approximately 130,000 users. Following the launch of ChatGPT, the company has observed a significant surge in sign-ups and platform usage, emphasizing its capability to facilitate collaborative AI agent development for diverse applications.

🫡 Meme of the day

⭐️ Generative AI image of the day