Welcome to AI Logs! We have a lot of exciting news and announcements in store for you this week. Let's kick things off with the big news about Perplexity AI, which has landed itself right in the middle of a major controversy.

Perplexity: The 'bullshit' app?

When it came out in 2022, Perplexity AI garnered a lot of fanboys (and girls). But there was also a big section of people who didn’t find it all that great. It’s a free search engine and can easily summarize articles from the web without the hassle of clicking on multiple links.

But in the scheme of things, it seems to have forgotten to abide by a key internet rule—Robots Exclusion Protocol—which is a standard used by websites to communicate with web crawlers about which parts of the site should not be peered into. Like paywalls. A Wired report has called Perplexity AI ‘a bullshit machine’ claiming that it’s been scraping the entire internet hell or high water with no regard for walls.

The report also claims that the chatbot is flopping at its main gig as well—that is “summarizing stories inaccurately and with minimal attribution.” It's hard to mess up models these days with the introduction of bigger and better LLMs. And on that note, we have a new king of AI models this week...

Anthropic releases its most ‘impressive’ model yet

No talk, all business. That’s Anthropic for you. The company, which is the arch-nemesis of OpenAI, unleashed its newest brainchild five days ago - Claude 3.5 Sonnet. It’s the first in line of its Claude 3.5 model series. If you look at this comparison chart they put out on their blog, Claude 3.5 Sonnet can run circles around its competition, leaving OpenAI’s GPT-4o, Google’s Gemini-1.5 Pro, and Meta’s Llama-400b eating the dust.

It's a pretty big improvement, considering its older sibling, Opus, was launched only three months ago. The 3.5 is 80% cheaper than Opus, has double the operating speed, and has a decent math and coding jump.

Additionally, Anthropic has introduced Artifacts - a feature that sits alongside Claude’s chat interface, allowing users to manipulate and refine AI-generated content in real time. More on this later in our ‘AI tutorial of the week’ section.

Crazy to think Anthropic is the company that delivers the most and on a consistent basis. Meanwhile, OpenAI is still holding out on us with their "coming weeks."

Speaking of OpenAI…

…tis’ officially the season to shop. It has reached that point in its life cycle where it will start gobbling up smaller companies. The cycle is surely starting to feel faster and faster.

In the last week alone, the Microsoft-backed company has purchased two companies. First is the real-time analytics database company Rockset through which OpenAI is looking to enhance the infrastructure of its enterprise products. Details of the purchase aren’t known, but the handover was reportedly in a nine-figure stock deal, making it OpenAI’s biggest buy ever.

The second buy is Multi (formerly Remotion), a remote multiplayer desktop app that allows more than one participant to connect at once. Does that mean we are looking at ChatGPT controlling your screen? I'm not sure how I feel about that, but my initial reaction is a bit of a queasy stomach.

Also, the ChatGPT desktop app for Macs is now available to all. It was once the exclusive playground of GPT Plus subscribers. There’s a bit of a catch, though. The desktop app only plays nice with Macs powered by Apple Silicon running macOS 14+ or newer. So, if you’re rocking an older Mac (like I do), that's too bad.

As OpenAI welcomes the new, it is also saying out with the old (*cough* Sutskever *cough*).

Sutskever’s new company

After months of playing hide-and-hide, Ilya Sutskever, the man behind ousting Sam Altman from their jointly co-founded company in November last year, has launched a company to develop safe superintelligence, which incidentally is also his company’s name.

As the founder and chief scientist of OpenAI, he tried to oust Altman over safety and risk prevention concerns. The attempt failed. However, this time Sutskever is really making sure that the world knows about his intentions to distance himself from profit-making models, especially after The Information reported that Altman said OpenAI could become a benefit corporation.

And before this newsletter gets too long, let’s do a quick wrap-up of the other important things that happened last week.

Dell and Nvidia are building an AI factory for Elon Musk's xAI

Dell CEO Michael Dell announced in a tweet that the company will be purchasing Nvidia chips to train Elon Musk’s chatbot Grok, a competitor to ChatGPT. Musk confirmed that Dell, along with Super Micro Computer, would be involved in assembling the supercomputer that his startup, xAI, is building.

McDonald’s pulls AI drive-throughs after order mix-ups

"Welcome to McDonald's, how can I serve you?"
"I would now like a Vanilla Soft Serve cone."
“Sure, I will add multiple stacks of butter to your order.”

Yep, that’s a real conversation that took place between a customer and its AI drive through. And after customers shared similar hilarious blunders online, McDonald's decided to yank its AI-powered ordering tech in the US. This system was developed by IBM with a voice recognition software, and was rolled out in 2019. However, instead of smoothly taking orders, it often misinterpreted what the customer wanted.

Meta tags real photos with ‘Made with AI’ tags

Photographers are complaining that Meta AI is tagging real photos with ‘Made with AI’ tags on Instagram, Facebook, and Threads. Meta announced this move in February and implemented it recently.

This also urges the photography community to question: How much auto-editing can be applied to a photo before it is considered AI-generated?

Apple, too, has announced that it would label AI-generated images in the metadata.

Major record labels sue AI over copyright ‘music swiping’

Sony Music, Universal Music Group, and Warner Records are suing two AI start-ups for alleged copyright violations in what could be a landmark case. The two AI music-generating companies Suno and Udio have been accused of copyright infringement on an "almost unimaginable scale." The lawsuit says these AI tools have software that "swipes" music to "spit out" similar tunes. If found guilty, the companies would have to pay a $150,000 compensation per tune.

In other news, Suno and Udio have gained attention for enabling users to record and upload their own audio to create music.

We have a social media platform where humans and AI interact

Social media is getting weirder. Butterflies is a platform where humans and AI personas exist together. The app’s interface resembles that of Instagram’s. How it’s different from Instagram is that you create an AI character called a Butterfly which generates photos and interacts with other accounts on its own. You can create as many Butterflies as you like, and they seamlessly coexist with human accounts, which can also post to the feed and comment. Is it weird or amazing? You tell me.

Another tool to try: Much like Perplexity AI and Google AI Overviews, Genspark uses generative AI to create custom summaries based on your search queries. For example, type in "How do I train my dog to do tricks?" and Genspark will generate a Sparkpage—a single-page overview compiled from various websites and online content. It’s like having a personal research assistant.

Did a friend forward this e-mail to you?

IE₊ SUPPORT INTERESTING ENGINEERING
_{Invest In Science And Engineering}

Enjoy exclusive access to the forefront of AI content, highlighting trends and news that shape the future. Join a community passionate about AI, delve into the latest AI breakthroughs, and be informed with our AI-focused weekly premium newsletters. With IE+, AI reporting goes beyond the ordinary - and it is Ad-Free.

NEWS

• AI-generated Elon Musk deepfake executes crypto scam on YouTube

  The Bitcoin scammers strike again on YouTube, posing as Elon Musk this time, asking viewers to hand over their crypto in exchange for double their investment.

• AI-backed ‘self-healing’ tech repairs electric grid automatically

  The system demonstrated near-optimal real-time performance which significantly reduced energy loss during outages caused by various scenarios.

• AI fakes Ukraine student into Russian women to lure buyers in China

  Olga Loiek’s image was manipulated by AI to create fake personas promoting Russia-China ties on Chinese social media.

• New AI detects 13 deadly cancers with 98% accuracy from tissue samples

  AI models can help detect early cancer by identifying specific DNA methylation signatures, leading to better patient outcomes.

MUST-READ

After OpenAI exit, Ilya Sutskever announces safe AI startup – SSI

Former OpenAI co-founder and chief scientist Ilya Sutskever has announced his new startup, which will focus on building safe superintelligence.

Sutskever has started Safe Superintelligence Inc (SSI) alongside his former OpenAI colleague, Daniel Levy, and Apple’s former AI lead, Daniel Gross.

According to a note on its website, the American company will have offices in Palo Alto and Tel Aviv, Israel, from which it aims to recruit technical talent.

AI PICTURE OF THE WEEK

We are doing things differently this week. Instead of discussing an AI image, we have a challenge for you. Below are two images. One of them has been generated using an AI tool. 

You have to guess which image has been generated by AI in the poll below.

TUTORIAL

This week, we will give you a breakdown of how to use Anthropic’s latest offering within its latest offering. Artifacts, a feature available in Claude 3.5 Sonnet, offers a separate window alongside your chatbox. This allows you to code, view, edit, and build projects in real time, providing a dedicated workspace for interacting with complex AI-generated content.

For instance, if you want to create a new version of the board game Chess and visualize its user interface for continuous improvement, Artifacts would facilitate that seamlessly.

Here's a breakdown of how to use this tool.

Prerequisites

• Access to Claude 3.5 Sonnet (available through Anthropic)

Activating Artifacts

Artifacts are typically enabled by default. However, if you're unsure:

1. In your Claude chat window, click the slider icon in the top right corner.

2. Select “Feature Preview” from the dropdown menu.

3. Look for an option titled "Artifacts" or "Enable Artifacts." If disabled, activate it.

Artifacts will be enabled for future conversations.

Using Artifacts

1. Prompt Claude for creative content: Ask Claude to generate things like code snippets, text documents, website designs, or even creative writing pieces. The more specific your prompt, the better the results.

2. Claude creates an Artifact: If the generated content is substantial (usually over 15 lines) and potentially reusable, Claude will create an Artifact for it. This appears in a separate window on the right side of the main chat.

Interact with the Artifact

The Artifact window allows you to:

1. View the content: See the full generated code, text, or design.

2. Edit and iterate: Modify the content directly in the Artifact window. Claude will update the Artifact based on your edits.

3. Version control: Access previous versions of the Artifact using the version selector at the bottom left.

4. Multi-window view: Use chat controls to open and view multiple Artifacts within the same conversation.

Exporting the Artifact

1. Copy content: Right-click and choose "Copy" to paste the Artifact content elsewhere.

2. Download file: If applicable, download the Artifact as a specific file type (e.g., HTML for a website design).