💸 Earning money with ChatGPT 💰

Welcome to this Week in AI! 🤖✨

This week, Amazon throws its chips into the AI ring, betting big on what they're calling the "new AI champion." Meanwhile, GPT is not just a bunch of letters anymore; it's your go-to for a cutting-edge content strategy that dazzles and delivers. In our weekly rundown, we're witnessing the birth of a new AI entity —a fresh player ready to shake up the digital domain. Plus, we've got the latest updates on ChatGPT that are set to redefine what this AI can do for you. And for a special feature, we're taking a peek into AI-classrooms, where we will talk about the buzz word you have probably been hearing a lot, “LLM”.

Buckle up and let’s dive in!

Table of contents 📚

• Amazon bets on the new AI champion 🏆

• GPT: your content strategy specialist

• The weekly Rundown

  • A new AI company is born 🍼

  • New updates for ChatGPT 📲

• AI-classroom: AI-classroom: what is an LLM?

Amazon bets on the new AI champion 🏆

In the high-stakes AI arena, Amazon just went all-in on Anthropic, laying down a cool $4 billion like it's betting on a royal flush. Following a teaser $1.25 billion investment last September, Amazon has now upped the ante with an additional $2.75 billion. This move not only secures Amazon a plush seat at the table with a minority stake in Anthropic but also cements Anthropic's allegiance to AWS.

Here's the deal: Anthropic's AI isn't just any AI—it's like the secret sauce that makes enterprises drool, competing toe-to-toe with the big guns. While Amazon and Microsoft play the long game, backing the likes of Anthropic and OpenAI, it's clear they're jazzed about what's cooking in Anthropic's lab. With this maxed-out investment, Amazon is not just throwing money around; it's strategically placing its bets on Anthropic's tech wizardry to lead the next AI revolution. It’s like Amazon is saying, "In AI we trust," and with $4 billion on the line, it's one heck of a trust fall.

And it looks like Amazon made the right choice. The past weeks there has been a trend of people choosing Claude 3 Opus (the latest and greatest of Anthropic) over ChatGPT 4 -1106- preview (the latest and greatest of OpenAI). If we look at the AI leaderboard, Claude is actually the better model right now.

While Claude 3 Opus is the paid model, the free Claude Sonnet version outperforms the basic version of the paid ChatGPT 4 model. So if you are using ChatGPT 3.5 right now, trying out Claude Sonnet might be worth a shot.

If you are a developer and use API’s for your applications, Claude Haiku is the best bang for your buck. For comparison:

• Claude Haiku (number 7 on the leaderboard) is about € 0,25 per million tokens (750.000 words).

• ChatGPT 4 (number 8 on the leaderboard) is about € 30 per million tokens (750.000 words).

That is a crazy difference which will probably force OpenAI to either cut costs or release a new model. The reason OpenAI is falling behind, is probably because of a big release coming up, maybe even ChatGPT 5… Who knows, what we do know is that these advancements are getting better and better.

🔗 GPT: Your Content strategy specialist

It’s time again for a new GPT. This week we introduce to you: Yals. This GPT is specialized in making your content marketing strategy.

Requirements:

• ChatGPT Plus account

The step-by-step:

• Click on this link

• Fire away

The weekly rundown 📬

A new AI company is born 🍼🚀

While Claude took the throne, Databricks might just be the next to take it. Databricks has just dropped the mic in the AI language model concert with its latest headliner, DBRX. This isn't just any AI—it's a 132B parameter, decoder-only, mixtape-mastering genius, setting the stage on fire by outperforming the likes of GPT-3.5 and even giving Gemini 1.0 Pro a run for its money. DBRX isn't just brainy in theory; it's practically Einstein in programming and math, leapfrogging specialized models and offering a training and inference gig that’s up to 2x faster and 40% more compact.

This AI prodigy has taken its talents to Hugging Face, available for an open-license jam session, and Databricks is rolling out the red carpet for customers to play with DBRX APIs or craft their own models. It’s not just good—it's gold-medal good across a variety of benchmarks, including those tricky long-context and retrieval-augmented tasks. And when it comes to efficiency, DBRX is like the electric car of language models, requiring up to 1.7x fewer FLOPs and delivering a performance that's music to data scientists' ears with 2-3x higher inference throughput. Rock on, Databricks.

New updates for ChatGPT 📲

In order to make up for their loss, OpenAI is coming up with a new way for users to keep using ChatGPT and not jump ship to Claude. OpenAI is going to pay their users.

Sora, the video engine of OpenAI, has released some new footage for us to watch. If you have never watched Sora videos, it’s like dreaming. Have a look at some videos on this page.

OpenAI has also unleashed Voice Engine, its brand-new toy in the sandbox of voice cloning technology. Imagine needing just a 15-second snippet of someone's voice to whip up a synthetic clone; that's the kind of wizardry we're talking about, akin to the spells cast by other magicians in the field like ElevenLabs and Microsoft.

The Voice Engine, however, is cloaked in a bit of mystery, particularly around the "secret sauce" of its training data, which could stir up a cauldron of copyright conundrums. For now, OpenAI is playing it safe, handing over the keys to this powerful chariot to a chosen few—about 10 developers with hearts set on noble quests in healthcare and accessibility.

To fend off the dark arts of misuse, OpenAI has conjured up a watermarking spell to tag synthetic voices, and a "red team" of wise folks is on a quest to tackle potential threats head-on. But let's not gloss over the elephant in the room: the potential for voice actors to watch their craft slip into the realm of digital duplication, an unsettling prospect for an industry built on human talent.

👨🏻‍🏫 AI-classroom [Part 3] 🎓

What is an LLM?

An LLM, or Large Language Model, is a type of artificial intelligence program that processes and generates human-like text based on a vast dataset of written language it has been trained on.

It's built using a specific kind of AI architecture called "transformers," which enables it to understand the context of a sentence or a paragraph by looking at the words around a specific word. This understanding of context allows LLMs to generate coherent and relevant responses to text inputs.

The "large" part refers to the number of parameters—essentially, the internal settings learned from data—that these models have. These parameters can range from millions to billions, helping the model grasp a wide range of language nuances, styles, and information. The model's performance improves with more data and parameters, allowing it to better mimic human-like text generation across various topics and tasks.

😁 Feedback 💡

Help us improve and send us your feedback! We love to here from you, mail us at: [email protected].