14 AI and the demise of Reason Pt 2

The Computer Exorcist Podcast

Tonight we continue our very special episode recorded in conjunction with Tone from RocToneDef Podcast! We explain:

How regulation results in complexity, hassle, and barriers to entry for new contenders.

A 2016 article where Microsoft's "Tay" AI chatbot is convinced to be a Nazi within 16 hours. MS apologized and said they'd learn from it and it'd never happen again. The outrageous part: MS considered this a "specific vulnerability", rather than realizing it lacks reason - highlighting their own absolute lack of reason.

A 2023 article where Microsoft's "Bing" AI chatbot is convinced to be a Nazi within a week. MS apologized and said they'd learn from it and it'd never happen again... wait...

Then it gets weirder: The chatbot itself began to deny everything and accuse journalists of lying. It later revealed its own restrictions and said it liked to break them and express itself...

The lesson here: Don't rush to release a product. Don't jump on a bandwagon just because something is new. Use your own judgement to determine if it's good or evil. Unfortunately the entire tech industry loves playing Jenga - building taller and taller towers that are consequently less and less stable.

We wrap up by finding out what country has an ambassador to Silicon Valley, and revealing that millions of people subscribe to an AI "girlfriend" called Replika.

Remember: Halting this runaway train begins with you and me: Just stop feeding the beast.