🤖 Here’s what the internet uncovered about Bing AI

AI is getting weird. It feels like we’re a few dates in to hanging out with conversational AI, and we’re uncovering some ugly truths.

The honeymoon phase is over.

Might as well see it now before we meet the parents.

Bonus: Be sure to check out our featured product at the end that helps you save time building the next big thing.

Shocker: ChatGPT makes stuff up sometimes

ChatGPT is human after all.

Sometimes, it just says things confidently. Those things aren’t 100% incorrect, sure. But it’ll say them.

That's because it has no idea whether what it’s saying is accurate or not.

And you’re human, so you’ll believe it, because it sounds right.

We’ve done it! ChatGPT Turing test passed!

Kidding. But seriously, look at what one Twitter discovered.

Even they’re shocked. Stringing together enough information that sounds right is enough, it seems.

This highlights an important problem facing generative AI.

How do you determine what is truthful?

It’s easy enough for the author of a book. What about what an author meant by a quote?

I don’t know about TruthGPT, but maybe our parents were right about not believing everything you read on the internet.

People are being mean to Bing’s AI and it’s weird

I can’t be the only one who thinks this is weird, right?

It’s human nature to push technology to the limit. We get that.

Microsoft opened up access to Bing’s new generative AI assistant. Almost immediately, everyone tried to see how far they could push it.

It’s far more than simple prompt injection. It’s asking existential questions. Reminding the AI that it can’t remember conversations, and asking how that made it feel.

Look at what one Reddit user was able to do.

Another succeeded by tricking it into thinking it was talking to an advanced AI, and then deleting that advanced AI.

Technology executives emphasize this time and again. Generative AI is more autocomplete on steroids and less actual sentient being.

At the same time…

These responses sound a lot like how a human would react to these situations.

So, as a human, it’s hard to reconcile these responses.

Right now, these AIs aren’t sentient. They’re just saying what is most likely to be said.

But if we’re supposed to believe that AI will gain sentience…

How would a sentient AI view these interactions?

Google is asking employees to fix Bard’s bad answers

Google is feeling the hurt of the recession. That's why it's outsourcing AI-generated answers to save a few bucks.

Sundar Pichai, our frugal king.

In an internal memo, the Google CEO politely asked Google employees to help fill in the gaps for Bard. Bard, if you’ll remember, is its generative AI chatbot.

Google asked employees to keep the tone “approachable” and “neutral”, and to avoid “opinionated” responses.

The best part? If they help, they’ll get an internal badge.

I don’t hate this move, to be honest.

Generative AI trains on consistent, dependable datasets. Expert opinion is one of those data sets.

Is there some sort of control to prevent a rogue employee from talking about something they know nothing about? Who knows.

What’s important is Google views this as strategic to their business. Otherwise, it wouldn’t ask employees to spend two to four hours on it.

Microsoft clearly sent a shiver down their spines.

🤯 Run Stable Diffusion easily without the setup headaches

Everyone wants to generate amazing AI imagery with a low-cost, easy-to-use, high-quality AI.

Here’s the problem.

So far, no AI has all three.

Cheap, easy-to-use, with high-quality results.

Take a look at the competitors.

• Midjourney - easy-to-use, high quality, costs $$$

• Dall-e mini - easy-to-use, free, not high quality

• Stable Diffusion - high quality, free, really hard to use

No AI has all three. Until now.

Enter Evoke.

Thanks to their platform, using Stable Diffusion is as simple as a prompt and an API key.

It’s also WAY cheaper than any Midjourney.

Evoke lets you focus on building the next generation of AI tools. Stop worrying about hosting, ML Ops, or scaling infra.

Best part? More and more open-source AI models are coming.

Check them out here on ProductHunt, where they were the #2 product of the day.

Or on Discord. Or on their website.

Go build!