Your AI News Brief for Friday, May 19:

-DeepBrain's AI Interviewer to screen candidates
-ChatGPT iOS App
-Apple bans AI tools
-Meta creates it's own AI Chip
-Plus, a new investment ETF was released that focuses entirely on generative AI

Here are the details (and videos) ↓ Meta AI introduces MTiA, an AI Chip that is up to 10 times faster than previous generations. MTiA will be used in Meta’s data centers to train and deploy AI models.

A ChatGPT model generated a 500% return in the stock market (trading options) over a 15 month period by assigning a sentiment score to news articles about publicly traded companies.

Research by University of Florida's Dept. of Finance ↓ A few things to note about this research:

- This research was tested on historical data; not carried out in real-time
- Only one method made 5x returns
- Assumed no transaction costs
- S&P500 had a -12% return in the same period

AI news never slows down.

Here's what happened today:

-AI CEOs meet at the White House
-Deepmind CEO says we'll have AGI in a few years
-Microsoft says AI is about to do some "real damage"
-Bing is open to everyone
-AI tool to extract data from PDFs

Here are the details ↓ Sundar Pichai (Google), Sam Altman (OpenAI), Satya Nadella (Microsoft), and Dario Amodei (Anthropic) met with VP Harris today on how to advance "responsible" AI.

https://twitter.com/i/status/1654283133376942081

Scientists have developed a decoder that can reconstruct continuous language from brain recordings (with no implant required) If you want to spend 15 hours lying perfectly still inside an MRI machine, you can start to train this non-invasive brain activity monitor to read your thoughts.

The results aren't perfect yet, but still impressive.

Decided to ask the alien what would happen if US banks were insolvent, per @balajis warning about the coming hyperinflation crisis.

Creative little LLM or a prescient forecast?

OpenAI is increasing the message cap daily in ChatGPT Plus

Here are 5 other ways to access GPT-4 ↓ Open Playground: Explore GPT-4 & more, an open playground for GPT models

nat.dev

Meta's new AI model LLaMA came out today.

The paper was a gentle reminder of how capable Google's 540-billion parameter model PaLM will be. This was the Massive Multitask Language Understanding test (Humanities, STEM, Social Sciences, etc)

When finetuned for the test, LLaMA omitted GPT's superior results from the chart because it is not of "moderate size"

In the not-too-distant future, anyone will be able to create any type of music they want via text prompts.

Here's a look into recent research on diffusion models for generating high quality music audio from text prompts:

(more examples below) "Noice2Music" is a project from Google Research that generates a short 30 second piece of music based on text inputs.

"Watermarking" AI-generated text.

Researchers from University of Maryland propose a way to discretely embed (and detect) special tokens in order to accurately determine whether or not text was generated from something like ChatGPT.

🧵↓ The watermark works by selecting a randomized set of “whitelist” tokens before any text is generated, and then softly promoting use of those tokens during sampling.

(a token can be a letter, word, numbers, punctuations or any other string that is useful for the specific task)

Lots of new AI projects come out every week, and some are better than others.

So here's 8 random AI tools that work as advertised:

↓ 1. Rewind AI

Makes anything that ever crossed your screen, or said out loud at your computer, searchable. Black Mirror vibes aside, you won't lose anything ever again.

rewind.ai

📽️Text-to-Video?

It could revolutionize entertainment as we know it.

Here's Phenaki, a model that can synthesize realistic videos from text prompt sequences.

More examples below ↓ Phenaki, developed by Google Research, compresses the video to a small representation of discrete tokens. This tokenizer uses causal attention in time, which allows it to work with variable-length videos.

🤯 Full body tracking now possible using only WiFi signals

A deep neural network maps the phase and amplitude of WiFi signals to UV coordinates within 24 human regions

The model can estimate the dense pose of multiple subjects by utilizing WiFi signals as the only input

🧵 -New research shows WiFi signals can serve as a ubiquitous substitute for RGB images for human sensing

-Illumination and occlusion have little effect on WiFi for interior monitoring

-They also protect individual's privacy and the required equipment is affordable

Digitally clone your own voice in 3 seconds: VALL-E

Inside the new language modeling approach for text-to-speech synthesis: The language model was trained on a huge amount of data (60,000 hours of English speech) and combines it with just 3 seconds of a person's voice and uses that to synthesize new, high-quality speech that sounds like the original speaker.

🎧 AI-read audiobooks are about to explode.

Apple is now allowing authors to convert their books into audio using text-to-speech, and the audio quality is insane:

Here's some samples from Apple. Details below ↓ Apple is allowing authors to use one of two companies to produce the AI audio:

▶️draft2digital.com

▶️lp.ingramcontent.com/publishers/aud…

Requirements ↓

Unlocking the power of ChatGPT:

If you're interested in exploring the full creative potential of ChatGPT, here's a list of 130+ creative prompts that showcase the power of AI chatbots.

Examples, links, and files below ↓ Prompts act as a starting point for a conversation with ChatGPT, and prime it for an ongoing conversation to assist the user in answering questions in a specific way to achieve the desired result.

Like everything else, there's an "art" to it.

How does GPT-3 perform on an actual IQ test?

UCLA published findings on how various Language Models performed on analytical reasoning questions against an average college applicant.

The AI was the clear winner. 🧵 Below are some extracts from the paper about GPT-3's performance:

(link to paper)
arxiv.org/abs/2212.09196…

Alright here it is...

15 of the best AI tools you can start using right now for:

⚪️ Writing
⚪️ Productivity
⚪️ Audio Creation
⚪️ Image Creation
⚪️ Image Editing
⚪️ Design

Cost and links included... Let's go! 👇 ⚪️ Open AI Playground ($0.02 per 750 words)
🔹 @OpenAI
🔹 openai.com

Text generation based on any prompt you write. Answers questions, writes copy, simplifies, and expands. Test out the powerful Davinci text model, or less costly legacy models in the "Playground".