💥 Amazon advances text-to-speech model

Amazon’s text-to-speech model (named BASE TTS) has been trained on 100,000 hours of public speech data and is one of the largest text-to-speech models.

Its size and the data (including English, Spanish, and Dutch speech data) used to train it have been contributors to its ability to offer natural-sounding speech.

Emotive sentences (eg. “Oh my gosh! That’s unbelievable!”) or foreign pronunciation (eg. “croissant”) that usually trip up text-to-speech models, sound natural with BASE TTS (listen here).

✅ Find exactly what you need: You can filter your job search by technical or non-technical roles, remote, hybrid, or office-based working, and salary expectations.

🫵 Set the cadence for new job notifications: Whether it’s hourly, daily, or weekly, you get to choose when you receive new job alerts.

⬆️ Get that dream job: Browse 2000+ AI-specific jobs from the top 100 AI companies, all carefully selected to ensure 100% relevancy and accuracy.

⌛ Save time and reduce effort: Browse, bookmark, and then apply to batches of jobs, all in one go, saving time and energy.

Microsoft has discovered that Chinese, Russian, Iranian, and North Korean hacker groups have been using Large Language Models (LLMs)—like ChatGPT—to improve their cyberattacks.

For example, Fancy Bear (linked to Russian military intelligence) has been using LLMs “to understand satellite communication protocols and radar imaging technologies.”

Other groups have been leveraging them for scripting, phishing, vulnerability research, detection evasion, and social engineering techniques, to optimize their hacking operations.

While Microsoft hasn’t detected any “significant attacks”, they’ve closed all accounts associated with the identified hacking groups to protect their users.

While the details remain largely unknown, it’s rumored that OpenAI is developing a web-based search engine—which could be integrated with ChatGPT—that’s expected to rival Google.

It will probably capitalize on Microsoft's backing and utilize Bing for some of its functionality, after all, ChatGPT Plus already offers users a ‘Browse with Bing’ feature.

The news comes after recent criticism of Google’s poor-quality search results (due to AI-generated content and black hat SEO tactics) and highlights a shift towards AI-powered search.

‘Arc Search’ (which browses the internet for you) and ‘Perplexity AI’ (which returns search-related text answers) are examples of AI search companies leading the change in web search habits.

Slack is launching AI features (to help workers find information quickly) including channel and thread summaries and the ability to answer Slack-content-related questions.

Users can ask it to outline topics discussed within a specific channel or summarize unread messages between custom date ranges within threads.

They can also ask it questions (similar to ChatGPT) and it will scour Slack to deliver answers, asking users for a quality check afterward to generate high-quality responses.

Although Slack AI is currently only available in English, for paying US and UK Enterprise users (cost is unknown), support for other languages and other plans is coming soon.

OpenAI’s innovations and internal progress often get leaked by people ‘on the inside’ and to combat this, OpenAI is seeking an insider investigator to “mitigate insider threats.”

For instance, recent insider reports revealed that data relating to custom GPTs (including initial setup instructions and customization files) had been exposed, causing havoc.

This new role will “involve conducting discreet investigations with HR, and shaping a secure culture through training” (for full details, check out this Work In AI job board posting).