Italy Unbans ChatGPT + Amazon Alexa LLM

OpenAI has restored access to its ChatGPT service in Italy after implementing changes to satisfy the Italian Data Protection Authority's concerns about unlawful data collection and the lack of prevention of underage users from accessing inappropriate material. The changes include a new form for EU users to remove personal data under the GDPR, a new tool to verify users' ages upon signup, and a help center article that outlines how OpenAI and ChatGPT collect personal information.

OpenAI may face further challenges as Spain, Canada, and other countries have opened or considered investigations into its practices, and European lawmakers are advancing the AI Act, which could add additional requirements for companies like OpenAI.

Now that Italy got OpenAI to budge on their privacy features do you think it's high time, other countries do the same?

Read More

Amazon’s Alexa LLM

Amazon CEO Andy Jassy announced during the company's first-quarter earnings call that the tech giant is building a more "generalized and capable" large language model (LLM) to power Alexa, which will be more effective due to the advent of Large Language Models and generative AI. Jassy believes that the addition of an improved LLM will help Amazon work towards its goal of building "the world's best personal assistant," and noted that the company has invested in AI and LLMs for years.

He didn’t give an exact date to expect the full launch but we can definitely expect something before the year runs out.

Read More

Microsoft New Text-To-Speech Model

TTS has a big problem, which is it only works well with a limited number of speakers in a recording studio.

Microsoft came up with a solution called NaturalSpeech 2 that can create better-sounding voices with more variety. They used a technique called "latent diffusion models" to train the program to make speech that's more expressive and can handle different voices and styles. They also used continuous vectors instead of discrete tokens to make the speech sound more natural.

Continuous vectors make the voice flow better, ever used those TTS where the voice fumbles at commas or doesn’t use “tone”?

In the future, they want to make the program even better at making singing voices, too. Who knows, maybe one day we'll have robots that can sing like Beyoncé but in the meantime, Text-to-speech technology is getting better at making fake people sound more like real people.

Read More

Paper

Not Major News but OpenAI Closes $29B Valuation

Dropbox is laying off 500 people and pivoting to AI

Prompt Injection Defense (SudoLang) (WIP)

This is a prompt security prompt, to prevent hackers from injecting your prompt for malicious purposes.

Works OK with Claude, and “good” with 3.5

If you’re building a prompt-wrapped AI tool, then this is worth taking a look at.

Prompt:

Copy/paste the prompts here: https://pastebin.com/7YaiZpeh

Prompt:

Copy/paste the prompt here: Click Here

Quiz Prompt for Students

Prompt:

Copy/paste the prompt here: https://pastebin.com/Ra6TrKCt

A good doggo cooking dinner

Doggo not dog