DAI#3 – AI sex, lies, and deepfake videotape

September 8, 2023

We had a bumper crop of AI news this week. Google marks a quarter century and is finally hitting top gear in the AI race. AI keeps getting smarter but apparently, you can hypnotize it into misbehaving.

Google is trying to stop politicians from lying but the Chinese and North Koreans aren’t helping. The AI race is hotting up and here come the Chinese!

Is old man Google waking up?

Google turned 25 this week and looks like it’s about to give those young AI startup whippersnappers a lesson in AI.

The company is even targeting the real tech old-timers. We got a preview of Duet AI for Google Workspace and the pricing is a shot across the bow of Microsoft’s 365 Copilot.

Google accountants will be watching that space but the really exciting stuff is happening over at DeepMind. Google’s Gemini model may be better than GPT-4. In an official response to the claim, OpenAI’s Sam Altman rolled his eyes and said, ‘Nah, ah!’

Is ‘deepfake’ just another word for ‘politician’?

It’s not like politicians need any help to lie, but AI certainly has supercharged their proclivity for fibbing. Google is trying to hold them back with its updated terms on political ads.

Deepfake technology is getting so good that it’s got us questioning everything. Was this Trump interview actually an AI fake? You be the judge.

via GIPHY

On the other side of the pond, the UK government is planning its AI Safety Summit. Its 5 objectives seem laudable but are they realistically achievable?

The global response to a virus that may or may not have escaped a lab was pretty poor. Do you trust them to keep you safe from what may be cooking in a computer lab somewhere?

The guest list for the summit hasn’t been released yet and we’re wondering if the folks on the other side of the AI Curtain will be invited.

The Chinese are coming!

Despite the US restrictions on chip exports, Chinese companies keep pumping out new AI models with more than 70 now available in China.

The new Chinese restrictions on generative AI seem to be having the desired effect. Baidu’s chatbot ERNIE appears to be toeing the line with some interesting responses to politically sensitive questions. We can’t see ChatGPT giving these kinds of responses anytime soon.

Chinese tech giant Tencent rolled out its new model called Hunyuan and they claim it’s better than GPT-3 and GPT-4.

Typically we’ve seen Western software solutions translated into Chinese for that market. Could we see more Chinese software translated into English and flowing West? Think it’s unlikely? You know who owns TikTok right?

While Google is trying to sanitize political ads, Microsoft released a report that says the Chinese and North Koreans are warming up their AI disinformation machines to influence Western elections. The translation of the tagline on Baidu’s page describes Ernie Bot as “useful, interesting and warm.”

via GIPHY

Look into my eyes. You are feeling naughty.

IBM researchers say that LLMs like ChatGPT can be ‘hypnotized’ into misbehaving. It’s really amazing how easy this was to do. AI is really good at lying and being sneaky.

Are current AI systems exhibiting any signs of consciousness? It’s an interesting debate and Sam’s deep dive covered the 14 “indicator properties” that a conscious AI would likely exhibit.

Microsoft says its Algorithm of Thoughts will make AI even smarter. This new approach to AI problem-solving could be a game changer in making models reason more like humans do.

As AI becomes more human, is it likely to become more or less trustworthy?

AI helps and hurts kids

OpenAI released a guide to help teachers use AI in the classroom but ChatGPT is also making things harder for educators. More prospective students are using ChatGPT to write their entrance essays making it tough to decide who to admit.

Adolescent mental health is on the decline and it probably has a lot to do with what the adults are up to. It turns out that AI is better at early predictions of self-harm and suicide than clinicians are.

While AI is helping kids in some ways, it poses serious threats in other areas. US state prosecutors are pressuring Congress to protect children from AI ‘sexploitation’. The explosion of sexual AI-generated ads and apps is a case in point.

Activision is trying to make gaming a safer space by using AI to clamp down on your toxic trash talk in Call of Duty. At least the kids won’t feel bullied while they kill each other online.

The media and AI – Can they work it out?

The Guardian joined a growing list of websites that are blocking OpenAI’s GPTBot from scraping their content.

It seems like a bit of a knee-jerk reaction because the long-term benefits of allowing the bot to do its thing don’t only lie with OpenAI.

Media organizations are forming a united front with a manifesto of Global Principles for Artificial Intelligence. The list of demands is understandable but it’s hard to imagine AI companies agreeing to them.

And they’re going to have to find a way to make it work. If these poorly written AI sports reports are anything to go by then AI reporters won’t be replacing humans anytime soon.

In other news…

Here are some of the other AI bits and pieces that got us clicking through:

And that’s a wrap of this week’s AI news. As usual, AI took us on an emotional rollercoaster ride. We got excited, a little nervous, and we’re still trying to work out why Trump sounded so weird in that interview.

There’s so much drama and intrigue in the AI industry that we nearly forgot that the actors and writers are still striking. With the battle between Google, OpenAI, Meta, and Elon Musk, AI may have already replaced them all with its own version of Game of Thrones.

For better or for worse, winter is coming.

Join The Future


SUBSCRIBE TODAY

Clear, concise, comprehensive. Get a grip on AI developments with DailyAI

Eugene van der Watt

Eugene comes from an electronic engineering background and loves all things tech. When he takes a break from consuming AI news you'll find him at the snooker table.

×
 
 

FREE PDF EXCLUSIVE
Stay Ahead with DailyAI


 

Sign up for our weekly newsletter and receive exclusive access to DailyAI's Latest eBook: 'Mastering AI Tools: Your 2024 Guide to Enhanced Productivity'.



 
 

*By subscribing to our newsletter you accept our Privacy Policy and our Terms and Conditions