All in on AI: OpenAI’s New Approach, More News

The Gist

AI safety. OpenAI says it emphasizes a multi-pronged approach to AI safety.
Say please to ChatGPT. People are calling for people to be polite to our chatbot friend.
Will Smith eats spaghetti with his hands. Yes, AI-generated images and videos continue to reach profound heights.

Real-world use and learning from mistakes are crucial for developing increasingly safe AI systems over time. Despite precautions, AI systems will always have unforeseen beneficial and harmful applications, necessitating continuous learning and adaptation.

OpenAI published a blog yesterday dedicated to explaining their approach to AI safety. Essentially, the team believes that as AI systems become more advanced, it’s essential to consider the potential risks associated with their use — something we were all reminded of with news of ChatGPT’s recent data breach, quickly followed by Italian security authorities issuing an order that prohibits the use of ChatGPT and prevents OpenAI from collecting information from Italian users.

In response, and to ensure that AI is developed and deployed in a safe and beneficial way, OpenAI has implemented a "multi-pronged approach" to AI safety that includes developing robust and verifiable AI systems, designing AI systems with human values in mind, creating a culture of transparency and collaboration around AI development and conducting research into the potential risks and benefits of advanced AI systems. Even with substantial safeguards in place with the release of new AI systems, sometimes nothing beats real-world training, mistakes and all.

“Despite extensive research and testing, we cannot predict all of the beneficial ways people will use our technology, nor all the ways people will abuse it,” company officials said in the blog. “That’s why we believe that learning from real-world use is a critical component of creating and releasing increasingly safe AI systems over time.”

In other AI news...

New AI Platform Aimed at Marketers, Social Media Managers

Pictory may be of interest to marketers and social media professionals. This newly released AI software tool enables users to easily create and edit high-quality branded videos, quickly and without any technical experience. Pictory utilizes AI to analyze a brand's existing video content and provide recommendations for video creation, including script writing and editing. Businesses can also customize their videos based on audience data and preferences, such as geographic location and social media activity.

According to the company, “Pictory has its sights set on helping marketers capitalize upon content languishing in their archives by repurposing it into engaging videos for sharing on social media and other marketing channels.”

Just Released: 2023 AI Index

The 2023 AI Index: A Year of Technical Achievement, Newfound Public Scrutiny, has been released and a few key trends emerged:

AI progress on benchmarks is slowing down, and new benchmarks are being developed to challenge advanced AI systems.
Generative models like DALL-E 2, Stable Diffusion, and ChatGPT are impressive but raise ethical concerns.
Large language models are becoming larger and more expensive.
AI is accelerating scientific progress and improving AI itself.

AI Video of the Week: Why Do You Want to Stop AI Innovation?

In the latest edition of Bloomberg Technology, hosts Caroline Hyde and Ed Ludlow interviewed DeepAI Founder Kevin Baragona, who talks about why he chose to sign the petition calling for a pause in AI development, and what he thinks the impact generative AI will have on the world.

Memory Lane: Will AI Help Us Remember Everything?

What if you could remember everything?

A new tech called “Ask Rewind” is promising just that — and has already raised more than $15 million from investors, including OpenAI’s Sam Altman.

Rewind Co-Founder and CEO Dan Siroker said he was inspired to create Rewind after he started to go deaf in his 20’s and then later received a hearing aide. If we have eye glasses to correct vision and aides allow us to hear, why, he wondered, can't we do something about memory loss? According to Siroker, 90% of our memories are forgotten within a week.

“Our vision is to give humans perfect memory,” Siroker said in a video posted to his company’s website.

Powered by OpenAI GPT-4, Ask Rewind is a downloadable search engine that allows you to record “anything you’ve seen, said, or heard” and then makes it searchable. Interested? Ask Rewind has a waiting list for anyone who wants to give the tech a try.

Spaghetti and AI: When Will Smith Goes Viral

I suppose that as an AI column, we have to at least mention Will Smith eating spaghetti — by hand.

Yes, that’s right, an AI-generated video of Will Smith eating spaghetti has been burning up social media. A Reddit user created the AI-generated video, shared it, and it soon spread across the Internet like wildfire.

It’s all thanks to ModelScope, by DAMO Vision Intelligence Lab, a research division of Alibaba.

This recently released, open-source AI tool is based on a “text2video" diffusion model and trained to create videos from prompts examining “millions of images and thousands of videos.”

As we can all see from the odd version of Smith throwing down a whole bunch of spaghetti, AI has yet to truly get certain aspects of the human form down — and the results are highly odd and pretty disturbing.