Would you confide in an AI life coach developed by Google?

Plus: 'Take Me to Church' singer Hozier considering strike over AI threat to music

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, guides, technical analysis and exciting developments in open source.

In today’s Dispatch:

  • (Long read) Fast Company dives into the search for intuitive user interface design in the era of generative AI. The article explores various startups and companies in Silicon Valley that are trying to build the next generation of interfaces to make AI accessible and useful in people's everyday lives. The field is still searching for breakthrough ideas to truly transform how we interact with AI.

  • Israel is aiming to become a global AI superpower. The country has launched the first stage of a National Artificial Intelligence program, and Prime Minister Netanyahu is considering a National AI Directorate. Meanwhile, Israeli companies like RiseUp and Aidoc are pioneering consumer and medical applications of AI technology. An index released in June ranks Israel 7th in the global AI ‘arms race’.

  • Experts are warning that deepfake detection tools being developed to combat false online content must be trained more effectively on different skin tones, or they risk being biased against minorities. Most current detectors rely on visual cues like blood flow and heart rate that don't work as well on darker skin. Without proper training, the tools could leave these groups more vulnerable to deepfake scams, fraud and misinformation.

Plus: How code-generating AI often increase, rather than decrease, the TCO of software (technical section), former Google CEO Eric Schmidt’s vision for marrying AI with science, trending tools and more!

Image: New York Times/Gabriel Alcala

From The New York Times (may be paywalled): In April, Google merged their legendary UK-based DeepMind team with Brain - the tech giant’s original Silicon Valley-based AI effort launched back in 2011. ‘Google DeepMind’ (the official post-merger team name) is now piloting a host of ambitious new tools for personal and professional assistance - even looking to turn generative AI into a life coach.

More details:

  • Google's new tools are being designed to manage at least 21 types of tasks, including sharing life advice, offering ideas, planning, and tutoring. Google has been integrating AI services into its professional suite, and is developing a tool to help journalists create more effective content.

  • Concerns were previously raised by Google's own A.I. safety experts regarding the potential dangers of users becoming overly attached to chatbots. Google has currently prohibited its chatbot Bard from offering medical, financial, or legal advice.

  • Google DeepMind has contracted a team from Scale AI, utilizing a team of over 100 data analysts and testers. Among other things, the workers are testing the assistant’s ability to answer intimate questions about challenges in people’s lives.

Takeaways: We’ve known for months that Google has been obsessively pushing AI into its suite. Google's latest endeavors reflect an overarching goal in the tech industry: the push for AI to be more deeply intertwined with our daily personal lives. While ‘cheerleader’-type AI tools and services are becoming more common, a meaningful AI life coach from Big Tech is uncharted territory. Given that some current concerns over using chatbots in healthcare would be relevant in this space, they will have to tread lightly.

Assuming development goes well, would you use/confide in an AI life coach from Google?

Login or Subscribe to participate in polls.

Photo: BBC Newsnight

From the BBC: Irish musician Hozier says he would ‘absolutely’ join potential strike actions against the threat of AI to the music industry, in line with similar concerns voiced by Hollywood actors and writers over AI's role in their profession.

More details:

  • AI's capability in the music domain extends to songwriting and mimicking known artists; this was highlighted when an AI-generated song with voices resembling Drake and The Weeknd went viral and now faces copyright issues.

  • Hozier stated: "Whether [AI is] art or not, I think, is nearly a philosophical debate. It can't create something based on a human experience. So I don't know if it meets the definition of art."

  • Google and Universal Music are reportedly negotiating to license artist melodies and voices for AI-generated music.

Takeaways: The progression of AI into artistic realms is igniting debates on the nature and definition of art, the boundaries of AI's role, and concerns over copyright infringements. Even potentially ‘good’ AI news in entertainment is often met with skepticism. Hozier's perspective that AI can't recreate art rooted in human experience resonates with the inherent nature of artistry that many believe cannot be replicated by algorithms. Still, if the music is compelling enough, people will listen to it - whether it’s art or not. That’s a tough pill to ask musicians to swallow.

Former Google CEO Eric Schmidt is building an ambitious new organization to tackle scientific challenges with the help of artificial intelligence.

Semafor • Reed Albergotti

Generative AI needs its pinch, its swipe, to be as big as the iPhone. Meet the San Francisco cool kids imagining your next computer as fluid as mercury, as malleable as snow.

Fast Company • Mark Wilson

More News & Opinion:

From our sponsors:

Fix Your Gut

Unlock a unique combo of personal insights and food and supplement recommendations to improve your digestive and overall health.

Schematics of the ChatGPT Chemistry Assistant workflow, with three different processes employing ChatGPT and ChemPrompt

From Chemistry World: Researchers at the University of California, Berkeley are using ChatGPT to help predict the best synthesis conditions for metal-organic frameworks (MOFs). Their approach speeds up the research process by removing tedious and lengthy human trial-and-error processes.

More details:

  • MOFs are a class of porous materials with a unique structure that creates an expansive surface area within a small space (often likened to several football fields of surface area on a single teaspoon). Their continued development is crucial to clean energy and transportation, drug development and other sectors.

  • The team utilized a series of specific prompts, enabling ChatGPT to efficiently mine and analyze data pertaining to MOF synthesis.

  • Instead of building new AI models from scratch which demands intricate coding knowledge, the team took advantage of the existing capabilities of ChatGPT.

  • MOFs can be made using a myriad of combinations of metal ions and organic linkers. However, the wide variety of synthesis methods and lack of a standardized format makes determining the best synthesis approach for specific combinations challenging.

  • The team used ChatGPT’s Code Interpreter to scan and process hundreds of papers related to MOF synthesis. They expect that AI models will eventually help create completely new materials to address problems relating to climate change.

Takeaways: There have been numerous applications of AI in fields where the vastness of data make it challenging for humans to parse and understand efficiently: protein folding, genome analysis, climate modeling, etc. While there is a justified media focus on the inaccuracies of current AI systems, they’re still already vital to advancing many fields of cutting-edge research. They just need (much, for now) human oversight.

To reduce total cost of ownership, choose the right abstractions. And if you are looking for generative AI to cut your TCO by writing code - think twice.

-

The New Stack • Laurent Doguin

More Open Source & Technical:

Social media/Video/Podcast:

  • (Discussion) AI can't take jobs soon enough! Faster, please! [Reddit]

  • (Discussion) ChatGPT holds ‘systemic’ left-wing bias, researchers say [Reddit]

  • This AI avatar just changed the world [YouTube]

  • Microsoft & OpenAI CEOs: Dawn of the AI Wars [YouTube]

  • This maglev train runs power-free with AI support [X]

Did you know? 

OpenAI's API pricing varies depending on the language used. Tokenization breaks down text into units, with the API charging based on these units. In English, the phrase "Hello world" is tokenized into just two units. However, when translated into languages like Japanese or Hindi, the number of tokens can increase dramatically, leading to higher costs.

The discrepancy comes from how different languages encode characters. While English characters generally require only one byte in UTF-8 encoding, a single Japanese character might need three bytes. Similarly, Hindi's Devanāgarī script can encompass multiple Unicode code points for one letter, each taking up three bytes.

Trending AI Tools & Services:

  • MealPractice: AI recipe generator - effortless (and random!) meal planning

  • DreamDecoder: see how far all-in-one AI website builders have come

  • MindPal: Dump all your files into a personal repository, your new second brain

  • FreeSubtitles.AI: Provides free audio transcription using AI

  • BeardStyleAI: Transform your look with 100+ AI-generated beard styles in minutes

Have a great weekend and we’ll see you Monday morning!

This is a total revolution that is coming. It’s going to totally transform almost all elements of life over the coming years, and indeed, even months, in some cases. It is much faster than other revolutions that we’ve seen and much more extensive, whether that’s the invention of the internal combustion engine or the Industrial Revolution.

UK Deputy Prime Minister Oliver Dowden, August 2023