- AI Weekly Wrap-Up
- Posts
- 8-16-2023
8-16-2023
Keith Collins MD & Addie Politi
August 16, 2023
Top Story
AI Has Already Created As Many Images As the first 150 Years of Photography
More than 15 billion images were created in text-to-image apps in the past year. It took photographers 150 years to reach the 15 billion image mark, from the first photographs in 1826 through 1975.
Currently, users are creating an average of 34 million images per day.
The fastest-growing product is Adobe Firefly, which reached 1 billion images generated in just 3 months.
Startup Midjourney has 15 million users.
Approximately 80% of the images (i.e. 12.590 billion) were created with software based on Stable Diffusion, which is open source.
Clash of the Titans
Microsoft Dresses Up ChatGPT for Business
Microsoft has just announced a business-friendly version of ChatGPT, with strong protections for data security.
Businesses can run ChatGPT on their own networks, with no connection necessary to OpenAI. They can then use ChatGPT with their own proprietary data without fear of leaks. Microsoft provides Azure for secure cloud computing, and businesses can use MSFT’s industry-leading productivity tools, now AI-enhanced with their AI Copilot upgrade. It’s a pretty compelling value prop - expect corporate IT departments to love it. Because, otherwise, Blackberry Research reports that 75% of IT organizations are implementing or considering a ban on AI in the workplace.
Amazon gives product reviews an AI upgrade
Amazon user-generated eviews are potentially a highly valuable source of information for the prospective buyer, but slogging through all those reviews is a chore. Now Amazon is using AI to summarize and highlight product reviews, streamlining the process of selecting the best products.
Fun News
AI helps stroke victim to walk with “high-tech trousers”
Researchers in the UK have demonstrated AI-enabled garments that use electrodes to stimulate muscles in a woman’s partially paralyzed leg to mimic the motions of her healthy leg, allowing her to walk.
An Iowa school district is using AI to ban books
Deciding which books to ban is hard work - think of all the books you have to read just to find the icky ones. So Iowa book-banners have devised a shortcut - they just ask a chatbot to identify books that have sexually explicit or politically controversial content. So far, the chatbot’s naughty list includes such salacious material as The Handmaid’s Tale by Margaret Atwood, Beloved by Toni Morrison, The Color Purple by Alice Walker, and I Know Why the Caged Bird Sings by Maya Angelou. Hmmm… what could be the common thread?
AI is building highly effective antibodies that humans can’t even imagine
At a lab in South London, “a machine learning algorithm designs antibodies to target specific diseases, and then automated robotic systems build and grow them in the lab, run tests, and feed the data back into the algorithm, all with limited human supervision.” Some of the AI’s designs are so novel, it is unlikely that a human would have thought of them.
These women warned us about AI
For years, women, and especially women of color, have been sounding alarms about biases baked into AI systems, biases derived from their training data. This training data, largely drawn from the internet, includes heavy doses of negative stereotypes about almost all demographics except young white men. How could anything go wrong?
Short Takes
AI app lets users text with Jesus - here
White House offers $Millions in prizes for hacker-thwarting software - here
DefCon hackers invited by AI companies to try to make AI go rogue - here
What if generative AI turned out to be a dud? - here
Researchers find that Generative AI enhances creativity - here
Thinking fast and slow in LLMs - here
“Only AI made it possible” for researchers to track UK wildlife - here
China publishes sweeping new AI rules - here
FEC may regulate AI in political ads - here
OpenAI teaches ChatGPT to moderate itself, automagically - here
Saudi Arabia and UAE buy thousands of Nvidia AI chips - here
AI systems are made to run on a $100 Orange Pi microprocessor - here
A simple guide to fine-tuning LLaMa 2 on your own data - here
Economics
The AI backlash grows, with increasing concern about disruption to jobs
A report by Goldman Sachs back in MARCH projected that AI will shake up more than 300 million jobs around the world. Premier consulting firm McKinsey, followed with a statement predicting that by 2030, 12 million Americans will have changed their job due to AI. All this is adding to public concern about the potential negative impacts of AI.
AI was predicted to turbo-charge the economy, now naysayers are throwing cold water on the hype
Contrarian economist Tyler Cowen pooh-poohs expectations that AI will produce massive economic expansion, noting that the industrial revolution - “the single most significant development ever in boosting living standards” - only produced 2.4% annual GDP growth during its boom years in the 19th century. This would mean a doubling of the real economy approximately every 30 years, which can be massive in the long run, but less so on the scale of human lifetimes.
AI is overloading data servers and companies are scrambling to find a new solution.
A new hurdle is emerging in the AI industry, there is not enough space. Companies are trying to rethink how data is stored and find a new, more efficient way to deal with data. Nick Patience, senior research analyst at 451 Research says “the overarching takeaways from our 2023 Global Trends in AI study is that data infrastructure will be a deciding factor in which organizations emerge as AI leaders.”
Reactions
:quality(70)/cloudfront-us-east-1.images.arcpublishing.com/archetype/JLAYWMHCHJBE3G7AJWDUKVXZRQ.jpg)
Department of Defense creates new AI task force Lima
The new AI task force, code named Lima is under the CDAO (Chief Digital AI Office), a unit that has only existed for a year. Kathleen Hicks, Deputy Defense Secretary, has signed off on this task force and is confident that the government can manage the growing potential benefits and threats of AI.
Certain unexpected moral issues arise with commercially available chatbots
Certain moral concerns about AI are widely discussed (Will they kill us? Enslave us? Take our jobs? Abet crime and disinformation?). Newer moral and ethical concerns have become more salient as AI developers create more complex systems. With AI chatbots already creating what users regard as ‘meaningful’ and ‘deep’ statements, users can become confused as to how much to trust commercial AI systems. Eric Schwitzgebel, a professor of philosophy at University of California, suggests developers “avoid creating AI systems whose sentience or moral standing is unclear, and AI systems should be designed so as to invite appropriate emotional responses in ordinary users.”
More on AI backlash from Forbes
Using a restaurant analogy, the article compares AI to a suspicious food. Will it deliver a nourishing repast, or leave us with a stomach ache and a hangover?
That's a wrap! More news next week.