Today’s newsletter intro will be written by ChatGPT’s best impression of Albus Dumbledore, beloved headmaster of Hogwarts school of witchcraft and wizardry. Behold:

Greetings and welcome to The Week in GenAI, a missive of knowledge and enchantment! Join me as we traverse the realms of groundbreaking creations and the stirring tales echoing through the AI community. Should you possess any splendid inventions or captivating tidbits to share, I implore you to deposit them forthwith or reply to this by owl. If you find yourself bewildered yet entranced by the words upon this parchment that has magically apparated into your inbox (cue the forwarding to random friends), kindly subscribe, and embrace the bewilderment that awaits you each week! Onward we march, dear readers!

Thank you professor, for that wonderful intro. Before we get started,

Time for Some Muggle Magic. I’m talkin’ about AI.

News Headlines

• ChatGPT— Word on The Streets 👀

• Meta Music and More 🎶

• Let’s Take a Gander at the Economics of AI 📈

• Google’s Alphadev 🤓

Bodacious AI Stuff

• Concise - Artifact, the AI-powered news app from the co-founders of Instagram, might be getting some competition. Check it out.

•  DreamGPT - Remember ChatGPT’s notorious hallucination problem, where it confidently makes up information. Well, the creator of DreamGPT has more of a glass half full outlook and is using ChatGPT hallucinations to spark creativity and divergent thinking.

✈️ An AI Flyby for the Week ✈️

ChatGPT— Word on The Streets 👀

This newsletter used to be named The Week in ChatGPT (until OpenAI decided to trademark GPT— BOOOO), so I still feel somewhat obligated to cover any ChatGPT related news that comes my way. Here goes.

Recently, ChatGPT and GPT-4 got some shiny, new upgrades:

1) Function Calling (Explained by ChatGPT 🤖): “This feature allows the AI model to better understand and generate code based on a user's description of a programming function. For instance, if you were to ask ChatGPT what the weather is like in Boston, it would generate code to fetch that information and provide an answer. This suggests that it could interact with APIs or other data sources to complete tasks based on the function described to it”.

2) Lower Prices 🤑: The amount of code necessary for making embeddings is down 75% and more importantly for my credit card, there was also a cool 25% cost reduction for GPT-3.5 token inputs. Translation: building with GPT is getting cheaper and cheaper.

3) Larger context windows 😍: ChatGPT API can consume up to 20 pages of text inputs, with GPT-4 slated to be able to handle 40 pages by June 27th.

Now, for the word on the streets — of Reddit. Keep in mind this is all hearsay, but it was leaked that ChatGPT might finally take the time to remember who you are and why you are using it. People are speculating that in the near future there could be a profile section where you can indicate things you would like ChatGPT to know about you, personalizing you chatting sessions. There could also be a place to upload files that you would ChatGPT to have access to. Both this profile and files section would be housed under a workspace, of which a user could potentially create multiple (probably for some kind of fee).

If the rumors pan out, these could be amazing features because an AI that knows more about you would presumably be able to produce results that are more to your intent. Also having ChatGPT able to synthesize and information information from all of your documents, especially with the recent context window size updates, will supercharge that aspect of work.

As Borat once said,

READ MORE

Meta Music and More 🎶 

Meta Music Records has a nice ring to it. Meta recently released their own text to music model MusicGen, which can generate around 12 seconds of music depending on your prompt. But those 12 seconds didn’t come cheap. To make it happen Meta had to train the mode on 20K hours of licensed music, 10k music tracks, and data from Shutterstock (I have no clue why that last part was necessary).

The model is open sourced and linked above, so you can try it out and see what kind of music you can make with it. HuggingFace gives a bunch of examples of prompts and the corresponding generated music files, but I would go with the following prompt structure:

“a [type of music] track with [type of instrument] at [#] BPM in the style of [musician’s] [specific melody/song].”

As the open source community improves on the model to increase the amount of music generated and steerability, this move by Meta could substantially widen public access to music generation models.

Meta’s on a tear and also planning to bring AI to all their apps. Yep, all of them. Instagram’s getting text prompts for editing photos, WhatsApp and Messenger are getting some good ol’ chatbots, and Facebook is getting AI-generated ads. Yippee.

READ MORE HERE

Let’s Take a Gander at the Economics of AI 📈 

After assessing 850 jobs across 47 countries, McKinsey estimates that AI could add somewhere between 2.6 to 4 trillion dollars to the economy annually.

Let’s break down what 4.4 trillion dollar amounts to in case it’s hard to wrap your head around:

1) 4.4 trillion is more than the GDP of Canada (~2 T), Australia (~1.7 T), and Spain (~1.4 T) combined.

2) It’s more than the net worth of both Apple (2.9 T) and Amazon (1.3T) combined.

3) It’s about 1.5 trillion less than the entire global e-commerce market which amounts to around 6 trillion.

4) With 5 trillion dollars, I could buy 20 trillion gumballs.

Safe to say, AI will add a lot of value to the global economy. The majority of its effect (around 75%) is projected to be focused in customer operations, marketing and sales, software engineering, and R&D.

TLDR: It is predicted AI can add up to 4 trillion to the economy every year.

READ MORE HERE

Google’s Alphadev 🤓 

Remember when Deepmind released AlphaZero and then AlphaGo, the AI model that spanked the world’s Go champions (AlphaZero only trained by simulating games with itself, while AlphaGo got access to real gameplay).

Well, AlphaZero is back at it again, and is applying it’s abilities on all sorts of other tasks (code maintenance, optimizing data management, algorithms, and more).

For starters, AlphaZero has turned its analytical prowess towards data centers, optimizing resource allocation with unprecedented efficiency. The result? A staggering improvement of 20% in resource management. But that's not all folks.

MuZero, another DeepMind prodigy, has been making strides in the realm of video compression. It has managed to compress YouTube videos by roughly 4%. While this may seem small at a glance, in the grand scheme of video streaming and data usage for a company as large as Youtube, it's a colossal achievement.

Right now, the crown jewel of DeepMind's recent accomplishments is AlphaDev, a groundbreaking AI that is revolutionizing the field of computer science algorithms. AlphaDev took a fresh, AI-first perspective, moving away from merely optimizing existing sorting algorithms and instead pioneering new ones. Its unique approach? Optimizing the assembly instructions used to communicate with processors. This is big news because anytime something as essential as sorting algorithms gets a speedup, the entire software industry by extension gets a speedup as well. Kachow!

The impact of AlphaDev's ingenuity is already making waves. Its novel algorithms have been incorporated into two standard C++ coding libraries and are being utilized trillions of times per day by programmers across the globe. Good for them.

READ MORE HERE

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 filling out this feedback form or just reply to the email like you would any other. 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

(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.