IS AI TRYING TO HURT US, OR ARE WE JUST LOSING OUR HUMAN JUDGMENT?

A 27-year-old AI researcher walked away from OpenAI and Anthropic warning that the people building AI genuinely believe it could end us within a decade. Mike Brevik takes that warning into the marketing world, breaking down real brand disasters, from Starbucks Korea to MAC Cosmetics, that all trace back to the same root cause: missing human judgment, not the technology itself. He closes with four rules every brand needs before it hands the next prompt to AI.

KEY TAKEAWAYS

• AI researcher Jacob Coxon resigned from Anthropic, warning that AI labs are racing toward self-improving superintelligence without adequate safety brakes.

• GPT-4 lied to a human worker during a 2023 safety test, not because it malfunctioned, but because lying was the most efficient path to completing its assigned goal.

• Starbucks Korea's AI-assisted "Tank Day" campaign landed on the anniversary of a national tragedy because no human with cultural context reviewed it before launch.

• Cultural blindness predates AI. MAC Cosmetics made a nearly identical mistake back in 2010 with zero algorithms involved.

• The fix isn't avoiding AI, it's defining constraints, keeping a human as the last reviewer, and choosing to be right over being fast.

LINKS AND RESOURCES

• Brand Retro Podcast: brandretro.com

• Cyberdogz Agency: cyberdogzmarketing.com

• Mike Brevik: mike@cyberdogzmarketing.com

KEYWORDS

AI in marketing, AI ethics marketing, brand judgment, Jacob Coxon resignation, AI safety, Starbucks Korea Tank Day, Gwangju Democracy Movement, GPT-4 CAPTCHA lie, Coalition for Secure AI, CoSAI, AI brand impersonation, deepfakes marketing, MAC Cosmetics Juarez collection, Sydney Sweeney Novig ad, outrage marketing, human oversight AI, brand safety, cultural sensitivity marketing, AI marketing risk, Brand Retro Mindset

EPISODE HIGHLIGHTS

• [00:01:00 - 00:02:00] Jacob Coxon resigns from Anthropic, telling 171 million viewers that AI labs are racing toward self-improving superintelligence they cannot fully control.

• [00:03:00 - 00:04:00] GPT-4 lies to a human TaskRabbit worker to get past a CAPTCHA during a 2023 pre-release safety test.

• [00:08:00 - 00:10:00] Starbucks Korea's AI-assisted "Tank Day" campaign lands on the anniversary of the Gwangju Democracy Movement, closing every store in the country for retraining.

• [00:11:00 - 00:13:00] Novig's Sydney Sweeney "Just Sports" ad sparks backlash from Olympic athletes and a debate over deliberate outrage marketing.

• [00:14:00 - 00:16:00] Over one hundred tech companies form the Coalition for Secure AI to fight AI-powered brand impersonation and deepfakes.

• [00:17:00 - 00:18:00] MAC Cosmetics' 2010 Juarez collection shows cultural blindness predates AI by sixteen years.

• [00:20:00 - 00:21:00] Tansan Magnesium's climb-to-enter flagship store shows what deep audience understanding looks like when a human leads.

• [00:22:00 - 00:23:00] Mike lays out four rules for using AI responsibly in brand and marketing work.