The Week in GenAI’s back in business after spending two weeks out of commission (sorry, everyone 😔). Here’s me making it up to you with a newsletter jam packed with important developments in the generative AI landscape.

But

Now, get ready for models left and right today

A Whole Lot of Headlines

• Code Interpreter, GPT-4 API, and Browsing Plugin Removed 🫨

• OpenAI Courting Strangers for Their Data 📊

• Claude v2 Hittin’ the Streets 🏎️

• Meta’s LLAMA 2 🦙

• The People That Make ChatGPT Tick Aren’t Happy ⏰

• Elon’s Done it Again (Should be no surprise at this point)

Bodacious AI Stuff

• Who Feels Like Getting Their Career Roased? 

  This nifty AI web app by 6figr is prepared to roast yours or anyone elses’ linkedin profile. As for the roaster, you can take your pick from bigwigs like Elon Musk to Donald Trump. For you career-oriented individuals, 6figr is also an AI powered job recommender and career advisor.

• Want to Take a Crack at the Fantasy Bestseller List?

  With this project, gpt-author, you can start generating an original fantasy novel. All you have to do is provide an initial prompt and enter how many chapters you'd like it to be, and the AI then generates an entire novel, outputting an EPUB file compatible with e-book readers. That’s insane. And as someone who loves fantasy and science fiction, I’m absolutely going to take a stab at this and see what I can make.

• A Marketplace Selling AI Prompts

  I haven’t seen an AI prompt marketplace in a while so I thought I would throw this one out there. It might help people find new ways to improve their prompting. And if you’re some kind of prompting expert, here’s a nifty way to potentially monetize that skill.

✈️ An AI Flyby for the Week ✈️

Code Interpreter, GPT-4 API, and Browsing Plugin Removed 🫨

All three of these in one headline is how you know it’s been a busy couple weeks for generative AI.

First, the ChatGPT Browsing Plugin was removed to the dismay of many. As for why it was taken down, anyone remember a couple weeks back when I mentioned that the browsing plugin had a handy way to get past paywalls and other ways of blocking content. Yeah, apparently, companies don’t like it when you can access their paid content for free, so OpenAI had to take down browsing promptly. It was that or get hit with more lawsuits, and this time the plaintiff wouldn’t be a comedian.

Essentially, RIP ChatGPT Browsing plug-in. In the slightly paraphrased words of Horace Slughorn (the potions master from Harry Potter), “Farewell ChatGPT Browsing, whose long and faithful friendship those who knew you won't forget! Though your body will decay, your spirit lingers on. May your descendants ever flourish and your human friends find solace for the loss they have sustained."

But OpenAI had the perfect stuff lined up to keep anyone from dwelling on the removal of the browsing plug-in: releasing both the GPT-4 API and Code Interpreter. There’s not much to cover about the API except for a price comparison to the ChatGPT API. It’s not completely apples to apples because they have different context window sizes, but between the 4K ChatGPT and 8K GPT-4, prompts are 20x times as expensive and outputs are 30x more expensive. Yowza, I know GPT-4 is better, but I don’t know if it’s that much better. I guess this is just what the top dog with the top model gets to do and everyone else has to pay up. But you could also look at this huge price hike as a strategy to squeeze as much profits as possible before competitors and open source catches up.

And now for code interpreter. You’ve most likely heard snippets of what it can do and the catchphrase, “Now, everyone can be a data analyst“ floating around. I’d revise it to now, “Anyone with a chatGPT premium subscription can be a data analyst now“, but yeah it’s pretty awesome. I’ll let Code Interpreter explain what it can do

Straight out of Code Interpreter:

1. Code Execution: The code interpreter can execute Python code and return the results in real-time. This is useful for running small pieces of code, doing calculations, generating plots, etc.

2. Data Handling: The interpreter can handle data transformations using Python's built-in capabilities as well as libraries like pandas and numpy. It can also visualize data using libraries like matplotlib and seaborn.

3. File Handling: The interpreter can read and write files. It can also handle a variety of file types, including .txt, .csv, .json, and others.

4. Code Debugging: The interpreter can help debug Python code by executing it line by line and identifying where errors occur.

5. Code Explanation: The interpreter can explain what a piece of Python code does, breaking it down line by line if necessary.

6. Code Generation: The interpreter can generate Python code based on specific instructions or requirements.

In case you missed it, this means you can now upload your data directly to ChatGPT instead of needing to use the API to bring ChatGPT to you. However, if you don’t want your data ending up in some gargantuan dataset, make sure you disable “Chat History & Training“ in your settings before you start using Code Interpreter (which you also need to enable in settings lol).

Ok I think I pretty much covered it. Only 6 more sections to go 😬.

READ MORE

OpenAI Courting Strangers for Their Data 📊  

"A model is only as good as its training data."

Basically, if you have garbage data, your model’s going to be garbage as well. OpenAI knows this and are looking past the expanses of the internet to the cordoned data stores of individual companies. If you’ve got the data goods, OpenAI wants a piece of that action (well, really they want the whole thing).

Recently, it formed two significant partnerships with The Associated Press (AP) and Shutterstock. OpenAI has gained access to AP's news archives dating back to 1985 and Shutterstock's repository of images, videos, music, and metadata. In return, these entities will have priority access to OpenAI's latest technologies.

The necessity of this approach is underscored by recent actions from major tech companies. For example, Google changed its data privacy agreement to train its AI model, Bard (it’s using all the public data on the internet now), while Stability AI faces a $1.8 trillion lawsuit from Getty Images for unauthorized data use. OpenAI itself had run into trouble trying to use data from Reddit and Twitter.

Data is the new "oil" for AI, and there is a significant shortage due to high demand. More data leads to better AI models, and there's a race to acquire the most unique datasets. OpenAI's partnerships will likely help improve their AI models' accuracy and the timeliness of the information they provide. The partnership with Shutterstock could also be particularly beneficial to the development of DALL-E, OpenAI's image generator.

And I expect to see many more of these partnerships forming in the near future. Each company will be a clue for where AI models are headed.

Claude v2 Hittin’ the Streets 🏎️

Anthropic recently launched Claude v2, an AI chatbot competitor to ChatGPT, in the US and UK. This model, while slightly trailing ChatGPT, boasts fresher training data (up to 2023), and is capable of handling much more context — up to 75,000 words (100K tokens), compared to ChatGPT's 3,000-word limit.

Another thing that sets Claude v2 apart from ChatGPT is that it’s less likely to generate dangerous or harmful content due to being trained on a "constitution" or rulebook. Claude is more likely to answer with a “I can’t help with that” instead of a harmful / incorrect response / hallucination. Speaking of an AI that’s more truthful, Elon’s coming back in a later section and he’s got more in store than just TruthGPT.

Back to Claude, some key points of this AI are:

• It outperforms GPT4 on the GRE writing, HumanEval coding, and Bar exam.

• It has abilities similar to ChatGPT Code Interpreter.

• It can import an entire PDF file and summarize the document.

• It's 5x cheaper than GPT-4 for developers.

In terms of competition, Claude v2 offers serious competition to ChatGPT. However, currently, ChatGPT retains the lead in terms of features and backing, since it’s got the burgeoning plugin ecosystem and that cool 13 billion from Microsoft (OpenAI’s valuation is ~30 billion while Anthropic is at 4 billion).

READ MORE HERE

Meta’s Makin More Llamas? 😱

Meta released its advanced AI model, LLaMA 2, to the public for research and commercial use. Unlike ChatGPT, LLaMA 2 is fully open, allowing anyone to build on it, and it performs as well as ChatGPT-3.5 on a slew of benchmarks (except coding), marking it as a serious competitor. This release has potential benefits for businesses, as it allows for the creation of customized models by adding data to the base model, which is not possible with ChatGPT.

Key points of the release include:

• Available for commercial use unless a product has more than 700 million monthly active users (Meta can’t have someone making the next Facebook using their model)

• It's available on Microsoft Azure, with plans for future availability on AWS and Hugging Face. You can try it out on Perplexity too if you’d like to play around with it.

• Meta has partnered with Qualcomm to make llama models run more efficiently on Qualcomm mobile chips. How much better will phones become with access to AI?

• Open-sourcing LLaMA 2 could disrupt the entire Large Language Models (LLM) market and challenge existing competitors and startups.

However, the release of LLaMA 2 also raises potential concerns about misuse and privacy. Meta has conducted safety evaluations and incorporated measures for safer use, and they've provided guidelines for responsible and open innovation. The overall impact of this move will depend on how effectively the associated risks are managed and how the technology is adopted and innovated upon by the global developer community.

READ MORE HERE

The People That Make ChatGPT Safer To Use Aren’t Happy ⏰

Kenyan workers who contributed to content moderation and AI efforts for OpenAI's ChatGPT are calling attention to their conditions. They've lodged a petition asking their government to look into the outsourcing practices of prominent tech firms in Kenya. These workers allege they suffered unfair treatment, lack of adequate support, and were exposed to distressing content that negatively affected their mental health. They're advocating for regulations that ensure the protection of workers in similar positions and prevent the outsourcing of potentially harmful technology without repercussions.

Sama, the company implicated in the dispute, services large clients such as Google and Microsoft. And Sama isn’t the only type of such platform that’s getting backlash. A study by the University of Oxford’s Internet Institute revealed that none of the 15 digital work platforms evaluated, including Amazon Mechanical Turk, Scale AI, and Appen, met fundamental fair work standards. The report highlighted that 250 workers from these platforms, commonly used by AI firms for gig work, dedicated nearly 27% of their time to unpaid tasks, resulting in an average hourly wage of $2.15 when taking unpaid time into account.

Back to Sama though, they have recently dismissed some workers but expressed willingness for audits to ascertain fair work conditions. Meanwhile, OpenAI, while recognizing the demanding nature of the work, asserts that it has established ethical and wellness standards and appreciates the efforts of these workers in enhancing the safety of AI systems.

READ MORE HERE

Elon’s Back in the News with the Launch of X 🚀

Elon Musk has unveiled his new AI venture, known as "xAI", which aims to "understand the true nature of the universe." Only a billionaire could get away with making their company’s mission statement this grandiose. Also, the amount that domain must’ve cost is probably around 6 figures a year. Classic Elon.

The company has a star-studded team consisting of researchers and AI engineers hailing from OpenAI, Google, and other top notch institutions. While specifics about xAI's operations remain unclear, it is speculated that the company could have a more open-source approach to AI development, in contrast to the somewhat guarded practices of its counterparts like OpenAI (after all this is part of Elon’s grudge match with OpenAI).

A particular focus for xAI could be AI that is capable of solving mathematical problems, a notion supported by an xAI team member's statement: “Solve math and understand the universe.” However, the exact functions of xAI remain largely speculative at this point.

In terms of competition, xAI is not expected to directly compete with other AI chats, instead focusing on deep scientific queries as opposed to consumer-oriented questions. There was a recent Twitter Space where more info on xAI was revealed:

1) xAI with have a chatbot

2) It’s focus is on deep scientific questions, not everyday chatting

3) It will have access to training data from Tesla and Twitter. The physical world data from Tesla will be invaluable for a system whose purpose is the pursuit of scientific questions and research. I wouldn’t be surprised if more of Elon’s companies also serve as data pipelines for xAI, like SpaceX and the BoringCompany. After all, xAI may be expected to play a role in helping these companies develop and improve their technologies in the not so distant future.

5) Tesla is beginning to build a supercomputer called Dojo to start building robotaxis, but in it’s spare time Dojo will also be helping xAI get off the ground. Oh, cool, a supercomputer will probably help for sure.

Parting is Such Sweet Sorrow (But Please Keep Reading LOL)

Well that's a wrap on the newsletter. Before I launch into my ending monologue, please take a moment to drag this email to your primary, so that you never miss it each week! Thank you!

Let me know how I did by replying to this email. If I did well and you're in the giving spirit please support the newsletter by buying me a coffee or donating to my gumroad page. Most of the money you donate to me will go back into purchasing AI products to try out and review on the newsletter. Bernie says it better:

Also if you have a product that you would like to be featured on the newsletter remember to add it here!

Lastly, if you'd like to refer a friend just tell them to visit chatgptnewsletter.com to subscribe! Or you could always just forward them the email without any subject header. I'm sure they they would love that.

Cool Things You Should Try/Buy From Every Past Newsletter 🤯

(Newsletter 1) ==> Consult domain experts in your browser with ExperAI

(Newsletter 2) ==> buy a talking book from Konjer

(Newsletter 3) ==> Pay for DoNotPay lol. Irony aside, it really is pretty cool.

(Newsletter 4) ==> Multion.AI - not your average burger buying extension.

(Newsletter 5) ==> Why settle for regular email when you could have Intellimail?

(Newsletter 6) ==> The AI app store where you can Cookup just about anything

(Newsletter 7) ==> A cute, fuzzy AI-powered online meeting summarizer: Otter.ai

(Newsletter 8) ==> Tired of taking hours to find the perfect online purchase? Getproduct.help’s got you covered

(Newsletter 9) ==> Turn your Scribbles into art with ScribbleDiffusion

(Newsletter 10) ==> Ever wanted a search engine for your entire online life? Rewind.ai

(Newsletter 11) ==> Your Online Image Studio 🎬: Clipdrop

(Newsletter 12) ==> Use Generative AI to create online courses on any topic you can think of. For example, you can create lesson plans about yourself before you get famous. Learn.xyz

(Newsletter 13) ==> Replace Siri with ChatGPT— HeyGPT!

(Newsletter 14) ==> Autonomous AI Agents in Your Browser— AgentGPT!

(Newsletter 15) ==> Create your own AI music with musicfy

(Newsletter 16) ==> Detect AI written content more reliably with GPTZero’s browser extension.

(Newsletter 17) ==> Another AI model playground, but more than just openAI models

(Newsletter 18) ==> Digitize your mind with You.ai 😵‍💫

(Newsletter 19) ==> Learn how to effectively use AI in your work and life with Maven

(Newsletter 19) ==> Super stylish generative AI website builder Dora. Explora whole new way to make websites.

(Newsletter 20) ==> A TODO list that does itself? Sign me Spellpage!

(Newsletter 21) ==> Making presentations just got easier again with Gamma

(Newsletter 23) ==> Making news concise with AI

(Newsletter 24) ==> Making cool QR codes with AI

(Today’s Newsletter) ==> Reply with what should go here!!!!