$1B for ScaleAI, Scarlett Johansson vs OpenAI, Alexa’s Upgrade, and TikTok Symphony

PLUS HOT AI Tools & Tutorials

May 24, 2024

Welcome to our weekly news roundup!

Today, we have a classic week for AI: a massive round for a promising startup with a young founder, a scandal about how AI can steal from us, and another impressive deal involving OpenAI. As you can see, we have a lot to discuss, so let's get started.

If you think that modern AI devices are not very good... Then firstly, you're right, and secondly, maybe you can do better? Follow this post to find out how to build your own smart glasses:
$20 AI Smart Glasses You Can Even Build Yourself
Anvar Abrarov
·
May 23
Read full story

This Creators’ AI Edition:

Featured Materials 🎟️
News of the week 🌍
Useful tools ⚒️
Weekly Guides 📕
AI Meme of the Week 🤡
AI Tweet of the Week 🐦
(Bonus) Materials 🎁

Featured Material 🎟️

Scale AI Raises $1B and Doubles Valuation to $13.8B

There was a lot of important news this week, but we chose this as the featured material. Because we had a few reasons. Scale AI is one of those startups rarely in the news but plays a huge role in the technological advancement. Besides, we like to follow the founders, and Scale AI CEO Alexandr Wang clearly deserves our mention.

Wang launched Scale AI in 2016 when he was just 19 years old. Five years later, at the age of 24, he became the youngest self-made billionaire in the world.

So, first, about what happened:

Scale AI, which provides data-labeling services to companies that want to train ML models, has raised a $1 billion Series F round from big-name investors that include Amazon and Meta. Scale AI is now valued at $13.8 billion. By comparison, the startup was valued at almost half as much at $7 billion during the previous round.

We have an addition among startups with valuations above $10 billion. Scale AI is gradually moving towards OpenAI. Sam Altman's company was last valued in January at $80 billion.

Accel led the Series F, which also led the Series A and participated in subsequent rounds. In addition to Amazon and Meta, Scale AI secured the support of many other investors. The full list would be very long, so I'll name a few important ones: Cisco, Intel, AMD, Nvidia, and Service Now. Yes, everyone believes in Scale AI.

What Does Scale AI Do?

As I mentioned, the startup provides labeled data for AI training. Major companies use its sources to power LLMs like ChatGPT and Gemini. Scale AI also offers tools to help businesses and governments customize these models for their specific needs. From a technical point of view, the startup has made many advancements in this field.

Here are a few of them, as cited by Scale AI itself:

• Autonomy Data Engine powered breakthroughs in L4 autonomy.

• Public Sector Data Engine has powered many US Department of Defense AI programs.

• Scale partnered with OpenAI in the first reinforcement learning experiments with human feedback on GPT-2 and scaled these techniques to InstructGPT and beyond.

Scale supplies data to power nearly every leading AI model, serving organizations like OpenAI, Meta, Microsoft, and more.

Now, Scale AI is tackling a major hurdle in the field: the ever-growing hunger for data that comes with more complex models. Their solution? Building a "data foundry" to ensure an abundance of data specifically designed to push the boundaries of AI.

Keep your mailbox updated with key knowledge & news from the AI industry

This new phase, fueled by a $1 billion funding round, focuses on three key principles:

Data Abundance: Instead of simply having enough data, the company aims to create a surplus of data specifically prepared for AI use.
Frontier Data: Not all data is created equal. The company plans to develop "frontier data" that challenges AI in new ways, pushing its capabilities towards complex reasoning, multimodality, and even creating intelligent agents.
Better Measurement: To ensure trust and widespread adoption, the company also plans to improve how AI performance is measured. This will be crucial for building confidence in the technology.

Scale AI emphasizes that achieving data abundance requires expertise across engineering, operations, and AI itself. But with the new funding there shouldn't be much of a problem with that.

If we draw parallels with other industries, Scale AI reminds me more and more of the ASML. It develops equipment to produce the best and highest performing chips on the market. And while everyone thinks that the IT sector is completely dependent on TSMC, it is ASML that plays a decisive role. And it makes you wonder what heights Alexander Wang and his startup might have reached.

Sharing is caring! Refer someone who recently started a learning journey in AI. Make them more productive and earn rewards!

Refer a friend

News Of The Week 🌍

OpenAI Partners with News Corp (And It's a Huge Deal)

Remember last week when I wrote about OpenAI finally making a deal with Reddit? At the time, I speculated that it was one of the biggest deals of its kind, as OpenAI got the largest data set to train its AI. Now forget it, we've got bigger fish to fry.

As reported by the WSJ (in this case, it's the most reliable source), News Corp struck a major content-licensing pact with OpenAI, aiming to cash in on a technology that promises to impact the news-publishing industry. The deal could be worth more than $250 million!

OpenAI would use content from News Corp’s news publications, including archives, to answer users’ queries and train its platform. To give you an idea of the scale, this includes The Wall Street Journal, New York Post, HarperCollins, The Sun, The Times, and many others. Terms of content-licensing agreements between publishers and OpenAI aren’t public, but the News Corp deal is among the biggest reached to date.

TikTok Boosts Its Ads Business With AI

Big news: one of the world's most popular social apps incorporates AI into its advertising business. TikTok announced that it is launching a new suite, "TikTok Symphony," for brands. The tools will help marketers write scripts, create videos, and improve current assets.

Users will have access to an AI-powered video generator called Symphony Creative Studio. The company claims that this tool can generate TikTok-ready videos with just a few inputs from an advertiser. The studio also offers brands ready-to-use videos for ad campaigns based on their TikTok Ads Manager resources or product information.

In addition, the company showed off the Symphony Assistant. This chatbot aims to help advertisers improve their campaigns by assisting them in creating and refining scripts and providing best practice recommendations. Advertisers can also ask it to show what's trending on TikTok or pitch a few ideas for promoting a new product.

Scarlett Johansson vs OpenAI

Finally we can talk about Hollywood stars as part of our newsletter! No, I'm serious. Scarlett Johansson's lawyers have some serious questions for OpenAI.

After OpenAI held a live demonstration of the voice last week, many observers compared its new chatbot to Johansson's voice in the 2013 Spike Jonze romantic sci-fi film Her. Johansson noted that she was shocked, angered, and in disbelief because she didn’t expect such an act from Sam Altman.

Interestingly, Sam Altman had previously approached Johansson to provide the voice for GPT-4o. According to the actress, the first time was nine months ago. Then, when Altman said her voice could be "comforting to people," Johansson declined the offer. And later, two days before the GPT-4o announcement, Altman called again and was again turned down.

A delicate situation. If the substantial similarity of the voice we could write off as a coincidence, the same similarity coupled with references to the actress so easily justified is already more difficult to explain. Now Johansson regards GPT-4o release as a personal insult, and her lawyers are already preparing a lawsuit.

Creators’ AI Subscription is the best gift for Creators diving into AI.

Give a gift subscription

Anthropic Scientists Map an LLM's Brain

Researchers at Anthropic have gone to great lengths to solve the "black box" problem. If you're not very deep into the topic, even the best developers (OpenAI, Google, Claude, etc.) building advanced LLMs can't tell you exactly how they work or why they give a particular answer.

But recently, Anthropic took a big step towards figuring it out.

Here, Anthropic shows that a particular feature activates words and images associated with the Golden Gate Bridge.

Using a technique called dictionary learning, Anthropic found a way to identify sets of neuron-like "nodes" in their LLM that the program associated with specific "features.” Once a particular feature had been identified, the researchers could manipulate it, "artificially amplifying or suppressing it to see how Claude's responses change," the researchers said.

I highly recommend reading Anthropic's full results here.

Amazon Plans to Give Alexa an AI Overhaul

According to CNBC, Amazon is busy overhauling Alexa with generative AI. This will be a new stage in developing the company's voice assistant with many new features and capabilities. If you’d like to try it out, Amazon is already testing elements of the upgraded Alexa in preview. You can enable it by telling your Echo device, “Alexa, let’s chat,” Amazon will notify you when you can access the preview.

But there's another important point on this topic. The new Alexa will run on a paid subscription, which you'll have to pay for separately from your Amazon Prime.

Useful Tools ⚒️

Browsebuddy – An AI co-pilot for e-commerce

Usermaven 2.0 – Future of web and product analytics powered by AI

StarSearch – AI Copilot, but for git history

Arrange – ChatGPT for your calendar

Findr 2.0 – AI search assistant for work. It’s designed to supercharge your work searches. The platform taps into your existing work apps like Gmail, Slack, and Notion to find information instantly. I wouldn't say it's a game-changing AI, but it will definitely make your work days a little more comfortable.