Welcome to The Upgrade

Welcome to the second edition of my weekly newsletter, focusing on the intersection of AI, media, and storytelling. A special welcome to my new readers from Harvard, The Guardian, the US State Department, AI startups, and many other organizations — you’re in good company!

In today’s issue:

• Case Study: AI + Obituaries = SNAFU

• The Week’s Top AI Stories

• The Big Think: A Deepfake World: What's After “Post-Truth”?

• ChatGPT Prompt Template: Brand Guidelines

Case Study: AI + Obituaries = SNAFU

Last week, Microsoft’s MSN news service published an AI-generated obit about former NBA player Brandon Hunter, titled “Useless at 42.” The MSN story was nonsensical, garbled, and riddled with factual errors. Far from an homage to the tragically deceased athlete, it made a mockery of his life and legacy.

“Throughout his NBA profession, he performed in 67 video games over two seasons,” read the virtually incomprehensible article.

Futurism broke the news of MSN’s AI experiment gone wrong, after which the article was removed — but not before reputational damage.

One X user summed the situation up eloquently in a post viewed over 2 million times: “AI should not be writing obituaries. Pay your damn writers ⁦@MSN.”

Microsoft has already been in hot water this year for AI-generated news content. Earlier this summer, an AI-generated travel article about Ottawa, Canada, recommended tourists visit a local food bank, advising: “Life is already difficult enough. Consider going into it on an empty stomach.”

Why has MSN struggled so much with AI-generated “reporting”? It fired its human journalists at the end of last year. The race to the bottom for “minimum viable content” is on. For Microsoft, which owns a 10% stake in ChatGPT, it does not appear to be going well.

The Takeaways:

• Do not use generative AI for sensitive subjects

• Always have human review built into any AI-powered editorial processes

• Pay real people to edit news, or risk the consequences

The Week’s Top AI Stories

Regulation and Policy

• Tech leaders agree on AI regulation but divided on how in Washington forum - The Guardian

• IBM CEO: “Regulate AI risk, not AI algorithms” - IBM (opinion)

• We need a public forum on AI, not a closed-door meeting in D.C. - Wired (opinion)

Ethics & Safety

• Tech ethics expert warns AI race will 'end in tragedy' if Washington doesn't act - Fox Business

• Safeguarding AI Is Up to Everyone - Scientific American (opinion)

• We don’t know how AI works - Vox (podcast)

Privacy & Security

• Facial recognition startup Clearview AI could change privacy as we know it - Marketplace

• Google Bard update reveals a more powerful AI – but it might scare privacy purists - TechRadar

Legal & Copyright

• Franzen, Grisham and Other Prominent Authors Sue OpenAI - The New York Times

• Meta tells court AI software does not violate author copyrights - Reuters

In the Workplace

• Most bosses want their future employees to have skills in AI - Fast Company

• AI could make side hustles a $1.4T industry, Morgan Stanley says - Business Insider

• AI expert is a hot new position in the freelance jobs market - NBC

• To catch a bot: the battle between AI and AI Detection Software - Wired

• Humanoid-robot CEO of drinks company says it doesn't have weekends and is 'always on 24/7'- Business Insider

AI Storytelling Tools

• OpenAI releases third version of DALL-E - The Verge

• OpenAI Hustles to Beat Google to Launch ‘Multimodal’ LLM - The Information

• Amazon upgrades Alexa for the ChatGPT era - WIRED

A Deepfake World: What's After “Post-Truth”?

After the 2016 US presidential election and the UK Brexit referendum, a common adage among journalists and academics was the arrival of the “post-truth era.” The consensus was that we were entering an epoch in which there was no longer an objective, agreed-upon reality to which society was tethered. Pundits and experts repeatedly blamed social media platforms for algorithms rewarding engagement regardless of sentiment (😃 = 😡) for a great polarization among the populace.¹ Oxford Dictionaries even named “post-truth" the 2016 Word of the Year, defining it as “circumstances in which objective facts are less influential in shaping public opinion than appeals to emotion and personal belief.”

This bizarre “post-truth” reality has been gradually normalized despite Americans seeing its crescendo during the January 6, 2021, Capitol riots, when extremists largely radicalized online converged to violently interfere with democratic processes. As we know, they nearly succeeded.

In our next election cycle, the creators of disinformation, propaganda, and conspiracy theories will have far greater weapons in their arsenals thanks to generative AI. The widespread availability of these next-gen creative tools represents a Pandora’s box for manipulating public opinion. It also signifies a watershed moment for society’s relationship with our shared reality depicted online.

You’ve probably seen the “Pope in a Puffy Coat” image earlier this year. The photorealistic AI-generated image was so convincing that many, including journalists, believed it was genuine. While fairly harmless and entertaining, the incident underscores the increasing difficulty in differentiating between authentic and AI-created or manipulated imagery in the age of AI.

Another similar example this year was the fake images of an explosion near the Pentagon in May, which briefly spooked the stock market.

After going viral, these incidents were debunked fairly quickly by experts and savvy observers alike. The tells? The small details. The Pope’s hands and the fence’s geometric anomalies were key clues. The scary part is that this tech isn’t even close to its potential. It’s just beginning. Better versions of AI image, audio, and video generators are released virtually every week.

So, what comes after "post-truth”? How will we discern what is real online anymore from what isn’t? Check out next week’s edition to learn what techniques may help us discern authentic media content from deepfakes!

_{1 (Research since then has shown a much more complex picture of online information bubbles and their causes and impacts. Source.)}

ChatGPT Prompt Template: Brand Guidelines

Would you like to see a YouTube tutorial on how to use these prompts? If so, please reply to this email: “YouTube” or “YT”

Thank you for reading! Interviews, tutorials, and more think pieces are coming soon…

If you liked this edition, please share it with someone who might enjoy reading it. 📧 

Don’t be shy—hit reply if you have thoughts or feedback. I’d love to connect with you!

Until next week,

Peter

Psst… Did someone forward this to you? Subscribe here!