🚔 Ex-Google engineer arrested for AI theft

JOB BOARD | PRO | PODCAST | TOOLS | ACCREDITATIONS

Welcome to AI Tool Report!

Thursday’s top story: Ex-Google engineer is arrested and charged with stealing Google’s AI trade secrets, to further AI advancements in China.

Over 937 people have already signed up for AI Professionals: The only accreditation you’ll ever need.

Why not join them? 

🌤️ This Morning on AI Tool Report

1. 🚔 Ex-Google engineer arrested for AI theft

2. ⚠️ Microsoft engineer warns FTC about Copilot

3. 💰 a16z expected to raise over $3B for AI funds

4. 🔍 Microsoft relaunches Bing Deep Search

5. ♊ Gemini releases Grammarly-like feature

Read Time: 5 minutes

FINANCE

Brought to you by YellowBrick

TRADE SECRETS

🚔 Ex-Google engineer arrested for AI theft

Our Report: Ex-Google software engineer—Linwei Ding—has been charged by US officials for stealing AI trade secrets from Google, while secretly working for two Chinese companies.

🔑 Key Points:

• Ding allegedly stole over 500 confidential files containing information about Google's supercomputing data center infrastructure, used to host and train large AI models.

• He was stealing this information for a year—while secretly working for two Chinese startups—and using it to further the progression of its AI technology development.

• Google, which has “strict safeguards to prevent the theft of trade secrets” alerted authorities after they found Ding was consistently uploading files to his personal account.

• If he’s found guilty, Ding could be charged with four counts of federal trade secret theft (with $250,000 in fines on each count) and could also face up to 10 years in prison.

🤨 Why you should care: The US and China have been engaged in a bitter trade battle for years, with both sides attempting to gain a competitive edge over the other, and US Attorney General (Merrick Garland) has stated that this is “the latest illustration of the lengths companies in China will go to, to steal American innovation".

In Partnership with Graft

Frustrated by the time-consuming IT processes?

This tedious work not only eats into your valuable time but also delays critical decision-making and innovation.

Now, your AI teammate can analyze tickets and draft responses enabling you to focus on strategic innovation and system resilience. Let your AI teammate do the work for you:

• Effortlessly categorize and summarize tickets, dramatically reducing the hours you spend trying to understand internal needs and sentiment.

• Generate and maintain up-to-date responses, documentation, and reports effortlessly, keeping your team informed and aligned.

• Instantly find information across your apps, eliminating the endless search and boosting your decision-making speed.

• Ask your AI teammate questions about your docs and get instant answers backed by the information in your docs.

The best part? Your AI Teammate adapts and learns from your specific knowledge and expertise. This trainable AI teammate is designed to complete work just as you would, aligning closely with your preferred methods and goals.

Try it

1. Xembly: Get an AI executive assistant and 400 hours extra every year ⏳

2. GenWebBuilder is for transforming images into stunning websites

3. AI VC is an AI tool designed for pitching to Venture Capitalists

4. LandingPage.fyi is for planning your landing page structure

5. Stardog allows you to have conversations with your data

SAFETY

⚠️ Microsoft engineer warns FTC about Copilot

• Microsoft engineer, Shane Jones, has written to the Federal Trade Commission (FTC) alerting them to safety concerns surrounding Copilot Designer, Microsoft’s image generator tool.

• When testing the tool—which is powered by OpenAI’s DALLE-3 engine—Jones found that it was generating harmful images, and alerted Microsoft, who failed to take him seriously.

• In his letter to the FTC, Jones states that, despite repeat warnings over the last three months, Microsoft continues to market it with "significant public safety risks."

• This comes after Google was forced to take down its own AI image generator when users found that it had created misleading and historically inaccurate images.

Prompt Engineering

Use the Prompt Engineer GPT by AI Tool report to 10x the below prompt.

✅ Email marketing

^{copy & paste ⬇️}

"I need a [type of email] that will convince my [ideal customer persona] to purchase my [product/service] by highlighting its unique benefits and addressing any potential objections.”

FUNDING & INVESTMENT

💰 a16z expected to raise over $3B for AI funds

• Venture capital giant Andreessen Horowitz (a16z) is aiming to raise $6.9 billion from investors for a new collection of funds, including two focused on AI projects and initiatives.

• Half will go to the firm’s fourth growth fund and the remaining half will go to an AI gaming fund and a fund to back AI startups in a field it’s calling ‘American Dynamism’.

• ‘American Dynamism’ includes companies that “address the most pressing issues” in the US, within sectors, such as aviation, defense, and manufacturing.

• This fundraising effort—expected to close in early April—aligns with their strategic focus on AI, after they invested over $200M in Mistral AI (French GPT rival) in December.

SEARCH

🔍 Microsoft relaunches Bing Deep Search

• Initially launched in December (but taken down due to unanticipated popularity led to poor performance), Microsoft has once again relaunched Bing Deep Search.

• The Deep Search feature utilizes Bing's existing search index and ranking systems with GPT-4 to provide more insightful and comprehensive search results.

• GPT-4 will analyze user intent and make search queries more comprehensive, which will surface highly relevant information that won’t show up in traditional search results.

• The new feature also provides users with a more transparent experience, with disclaimers warning them that their "GPT-powered results take ~30 secs.”

PROMPT ENGINEERING

♊ Gemini releases Grammarly-like feature

• Google’s Gemini (formerly known as Bard) has released a new feature that allows users to modify the responses they receive, with users likening it to Grammarly (AI writing assistant).

• Before, if users weren’t happy with a Gemini response, they’d have to retype their prompt and Gemini would then regenerate an entirely new response to it.

• Now, users can select a particular part of the response and Gemini will display a pencil icon with four options (regenerate, shorter, longer, remove) and modify that part, accordingly

• This feature is available for English-language users and aims to enhance the user experience by providing more control over the generated text they receive.

🌐 The world’s first AI (from 1983)

We read your emails, comments, and poll replies daily.

Hit reply and let us know what you want more of!

Until next time, Martin & Liam.