Misinformation, Mistakes, and the Pope in a Balloon: What Rapidly Evolving AI Can and Can’t Do | artificial intelligence (AI)

GRAMEnerative AI, which includes large language models like GPT-4 and image generators like DALL-E, Midjourney, and Stable Diffusion, is riding a “storm of hype and fear,” as some commentators have observed.

Recent advances in artificial intelligence have prompted warnings that the rapidly developing technology may result in “increasingly powerful digital minds that no one, not even their creators, can reliably understand, predict, or control.”

That’s according to an open letter signed by more than 1,000 AI experts, researchers and funders, calling for an immediate pause on the creation of “giant” AIs for six months so that security protocols can be developed to mitigate their dangers.

But what is technology currently capable of doing?

You can generate photorealistic images.

Midjourney creates images from text descriptions. It has improved significantly in recent iterations, with version five capable of producing photorealistic images.

These include the fake images of Trump’s arrest, which were created by Eliot Higgins, founder of the Bellingcat investigative journalism network.

Midjourney was also used to generate the viral image of Pope Francis in a Balenciaga puffer jacket, which has been described by a web culture writer. ryan broderick as “the first real case of AI disinformation on a mass level.” (The creator of the image has said he came up with the idea after taking magic mushrooms).

Image generators have raised serious ethical concerns over artistic ownership and copyright, with evidence that some AI programs have trained on millions of images online without permission or payment, leading to class action lawsuits.

Tools have been developed to protect artistic works from the use of AI, such as Glazewho uses a cloaking technique that prevents an image generator from being able to accurately replicate the style in a work of art.

You can convincingly replicate people’s voices.

AI-generated voices can be trained to sound like specific people, accurately enough to fool a voice identification system used by the Australian government, Guardian Australia research has revealed.

In Latin America, voice actors have reported losing work because they were replaced by AI dubbing software. “An increasingly popular option for voice actors is to take low-paying recording jobs at AI voice-over companies, training the very technology that aims to supplant them,” a Rest of the world report found.

you can type

GPT-4, the most powerful model released by OpenAI, can code in all computer programming languages ​​and write essays and books. The great linguistic models have led to a rise of e-books written by AI for sale on Amazon. Some media, such as CNET, have allegedly used AI to write articles.

Video AI is getting a lot better

There are now text-to-video generators available that, as the name suggests, can turn a text description into a moving image.

Can convert 2D images to 3D

The AI ​​is also getting better at turning 2D still images into 3D visualizations.

Make factual errors and hallucinate

AI, particularly the large language models used for chatbots like ChatGPT, is known for doing errors of fact that are easily overlooked because they seem reasonably convincing.

For every example of a working use of AI chatbots, there seems to be a counterexample of their failure.

Professor Ethan Mollick of the Wharton School of the University of Pennsylvania, for example, tested GPT-4 and was able to provide a fair peer review of a research paper as if he were an economic sociologist.

However, Robin Bauwens, an assistant professor at Tilburg University in the Netherlands, was turned down by a reviewer for an academic paper, which probably had used AI, as the reviewer suggested that he familiarize himself with academic papers that had been invented.

The question of why AI generates fake academic papers relates to how large language models work: they are probabilistic, in the sense that they map probability onto sequences of words. As Dr. David Smerdon of the University of Queensland says: “Given the beginning of a sentence, it will try to guess the most likely words that will come next.”

In February, Bing released a pre-recorded demo of its AI. As the software engineer Dmitri Brereton As noted, the AI ​​was asked to generate a five-day itinerary for Mexico City. Of five descriptions of suggested nightlife options, four were inaccurate, Brereton found. By summarizing figures from a financial report, Brereton discovered, he, too, managed to falsify the figures.

Can create instructions and recipes (cursed)

ChatGPT has been used to write crochet patterns, resulting in hilariously darn results.

GPT-4, the latest iteration of the AI ​​behind the chatbot, can also provide recipe suggestions based on a photograph of the contents of your fridge. I tried this with various images of the refrigerator detective subredditbut not once did it return recipe suggestions that contained ingredients that were actually in the fridge pictures.

It can act as an assistant to perform administrative tasks.

“Advances in AI will allow the creation of a personal agent”, Bill Gates wrote this week. “Think of it like a personal digital assistant: it will see your latest emails, it will know the meetings you attend, it will read what you read, and it will read the things you don’t want to bother with.”

“This will improve your work on the tasks you want to do and free you from the ones you don’t want to do.”

For years, Google Assistant’s AI has been able to make restaurant reservations through phone calls.

OpenAI has now enabled plugins for GPT-4, allowing you to search the web for data and order groceries.