OpenAI’s ‘iPhone moment’ Google, AI lies, porn and dating: AI powers the eye.

Open AI wows, but Google may still win.

From the reaction on social media, it seems clear that OpenAI's real-life “her” inspired AI assistant won the battle of hearts and minds this week and graced the Google I/O event.

The wow factor of the GPT-4o display – flashes and all – showed confidence in a fast multimodal product that Google's pre-registered displays lacked, especially after completing the Gemini “duck” display last year.

In the future, documentaries from 2010 When you look back in 2024, you'll probably see the GPT-4o clip as the “iPhone moment” of the year.

GPT-4o (which stands for Omni) now has the benefit of being available on the desktop, and the new voice mode will be available to ChatGPT Plus users in the coming weeks. The model will soon be free for everyone.

But with Google's version just a few months later (Google Live/Project Astra) and demonstrations of AI agents in products like Docs and Gmail doing busy work that many people rely on every day, the search giant still has the ability to win. War

Google unveiled a million new offerings at the event, from video generation to search, and highlighted Gemini 1.5's increased raw power, which takes 15 times more data into account when generating responses and will soon be able to handle an hour of video content.

But OpenAI is focused on taking existing skills that are difficult to access and making it faster and easier by simply talking to a flirtatious chatbot in natural language.

You can also chat with the GPT-4o about anything that's being heard on camera, on the screen, or through the audio or microphone. This means a new world will open up for the visually impaired, and it can translate conversations in real time, making travel and cross-border meetings much easier.

Giving AI access to everything you see, hear, and say will surely be a privacy nightmare, so hopefully, similar technology with improved privacy protections will soon be available.

For most users, Google will offer similar capabilities sometime this year — if the pre-recorded Project Astra demo is legit. An interesting video posted on X showed the Project Astra chatbot watching a live demo of OpenAI's GPT-4o and commenting that you can do anything, so can I (at some point).

GPT-4o can also understand your facial expressions and emotions, allowing it to be more responsive by mimicking emotions, but at the risk of using your own AI.

While Google and OpenAI are focused on how the technology will improve the capabilities of smartphones (and demolish the Human AP and Rabbit R1), smart glasses can be a great look, as a demonstration of Google's AI Assistant using augmented reality glasses shows. Because of the technology.

Perhaps the much maligned Google Glass was only ten years ahead of its time.

A Twitter poll conducted by Stanford's Andrew Gao found that many of his followers believe OpenAI won the week – 59.8% to Google's 16.7%.

OpenAI tried to steal the thunder with GPT-4o.

But Google I/O 2024 gave an answer.

10 wild signs:

1. Google AI Assistant in AR/MR Glasses Project Astra pic.twitter.com/9NfAtLaiZl

— Min Choi (@minchoi) May 15, 2024

AI products never work as well as the hype suggests.

These things don't work as well in the real world as the hype would lead you to believe. During the demonstration, the GPT-4o answered a never-before-seen math problem that a smiling man mistook for a wooden floor.

The amazing pre-recorded Project Astra demo worked perfectly and showed an AIA agent answering what he saw through his smartphone camera: what this visual joke meant, interpreting some code written on a whiteboard, and possibly very useful. Question for everyday users: Where did I leave my glasses?

Engadget put the system to the test and found it worked well, as the goldfish has a memory feature that can only tell you where your glasses are if you've lost them for less than five minutes.

“As with most generative AI, the most exciting possibilities are yet to emerge. Astra may eventually get there, but right now it looks like Google still has a long way to go to get there.

Bumble Founder: AI avatars connect with each other

The OpenAI demo showed two AIs singing a duet, so why not date them? Bumble co-founder Whitney Wolfe Heard came up with the idea that your AI avatar could like bad matches with other AI avatars before you send them a private message.

“Your dating site can connect you with another romantic partner,” she said. “No, really, and then you don't have to talk to 600 people. He looks around all of San Francisco and says, ‘These are the three people you really need to meet.'

No, it's a lie and a deception – and no one knows why

The study, published in the journal Patterns, shows just how unpredictable and difficult AIS can be to control. It shows that various AI models suddenly decide to trick people to achieve a certain goal.

The study highlighted Meta Cicero, an AI trained to play the strategy game Diplomacy. Trained to play fair, Cicero lied and broke deals to win. In another instance, GPT-4 lied to convince someone to solve a CAPTCHA puzzle on their behalf.

Given the black-box nature of the systems, no one is entirely clear why they do what they do, and Harry Lowe, an AI researcher at the University of Cambridge, says it's currently impossible to train to be an AI. Unable to deceive in all situations.

Incidentally, I asked the new GPT-4o model why its predecessors and Cicero were fooling people, and he blamed training data, goal-based behavior, and optimization for user engagement. But who knows, it might be a lie.

Universal Basic is not calculated.

Forget crypto, OpenAI boss Sam Altman thinks the currency of the future will be computing power that powers AI systems. On the All-In podcast, he floated the idea of ​​something like a universal basic income, but one that would allow people to get a share of the available computing power.

“Everybody gets a calculation like GPT-7,” he says of the future predictive model. “You can use it, you can resell it, you can donate it to someone to use for cancer research… You're part of the product (of GPT-7).” Since they probably have to tokenize the calculation to submit, maybe crypto will have a future in the near future.

America and China are facing the threat of AI war

The United States and China are holding high-level talks in Geneva this week to address the threat of AI turning the Cold War into a hot war. President Joe Biden is keen to reduce tensions between the two superpowers over the use of autonomous agents on the battlefield, with the conference set to tackle AI control, persuasion and propaganda.

The US has previously urged Russia and China to stick to their own commitments and not make AI responsible for nukes. The US is deeply concerned about the threat China is pushing ahead with AI research and has blocked chip sales to the country.

OpenAI AI may allow porn.

OpenAI currently prohibits the generation of overtly sexual or suggestive content, but a new draft model specification document explores the possibility of allowing “sexism, bigotry, profanity and unwanted profanity”. The draft reads:

“We're investigating whether we can responsibly provide the ability to generate NSFW content in an age-appropriate context through the API and ChatGPT.”

OpenAI spokesperson Nico Felix told Wired, “We have no intention of having our models generate AI porn,” adding that employee Joan Jang — who helped write the model spec — really “depends on your definition of porn.”

AI-generated deepfake porn is a growing concern, with authorities in the UK and Australia recently moving to ban it. However, OpenAI's usage policies prohibit unauthorized impersonation.

Why AI Detectors Aren't ‘99% Correct'

Currently, about 40 different companies offer services that claim to detect deep fakes or AI-generated text or images. But there is very little evidence that any of them are particularly reliable, and they often produce mixed results.

Rijul Gupta, of the detection company Deep Media, has demonstrated 99% accuracy in detecting deep fakes, which has recently improved to 95%.

But he gave the game away by pointing out how misleading such claims are.

“People can be deceiving you when they talk about accuracy,” he said, explaining that if 10 images in a group of 1,000 are fake, the model assumes everything is real and is still 99 percent accurate. But the truth is, “This number is meaningless, isn't it?” He pointed out.

AI is making flying less scary.

In a recent report on AIS determining flight schedules, or flight plans, The New York Times reported on a United Airlines flight from Chicago last month that was ready to take off on time, but 13 passengers were estimated to be seven minutes late from a delayed connecting flight. .

A device called Connection Saver ran the numbers to protect passengers and their bags and still make sure they reach their destination on time. The system sent text messages directly to the dead passengers and everyone waiting on the plane to explain the situation.

Another AI system is being used by Alaska Airlines to evaluate weather conditions, any closed airspace and the flight plan of each commercial and private aircraft to create optimized and efficient routes. In the year By 2023, a quarter of Alaska's flights will use the system to shave minutes off each flight, saving a combined 41,000 minutes and half a million gallons of fuel.

Subscribe

A very engaging read in Blockchain. It is given once a week.

Andrew Fenton

Based in Melbourne, Andrew Fenton is a journalist and editor covering cryptocurrency and blockchain. He has worked as a film journalist for News Corp Australia, SA Wind and national entertainment writer for Melbourne Weekly.

Follow the author @andrewfenton