Where will the AI cash flow ? GetInference AI nuggets Jan 10

Hello pioneers ! It's been a wild week in the world of AI (again), but don't worry, I won't bore you for too long with the same old news you've already heard. I'm going straight for the good stuff, folks. My four predictions for where all the AI cash is going to flow for marketing startups! Alright, let's cut through the chit-chat and get straight to the juicy bits.

SPOTLIGHT

Looks like Microsoft is throwing a party and inviting OpenAI to be the guest of honor, with plans to fund the AI lab and possibly unveil GPT-4 as the entertainment. Rumors also suggest that they plan to use OpenAI's ChatGPT technology as a karaoke machine for their Office apps. Word on the street is that it'll let users belt out natural language prompts like a pro in Outlook, Word, and PowerPoint.

But honestly, I don't know about you, but I'm not convinced about using ChatGPT in Microsoft Word. It's like trying to make a gourmet meal with a toaster oven, sure it's technically possible, but you might get some unexpected results. And let's not forget the highly anticipated launch of ChatGPT in Bing in March, it might shake things up like a game of musical chairs.

In other news, OpenAI is currently in negotiations with Thrive Capital and Founders Fund, where existing shareholders can sell their shares for a company value of $29 billion.

All in all, these developments in Generative A.I. space since ChatGPT made its grand debut are big deal.

But what's really exciting for us marketers and creatives is how it'll affect our lives, our jobs. Some might say that A.I. will cause a ruckus in the marketing world in 2023, leaving some agencies obsolete, while others will thrive. Like a game of survivor but with AI startups, some will be voted off the island while others will be awarded a million dollars and a shiny new career.(link)

"Don't worry, AI won't replace you... it's the person behind the AI that will.

But jokes aside, it's important to start learning how to use artificial intelligence in your work and personal life. And I'm not just talking about Generative AI, the tip of the iceberg, I'm also talking about supervised learning for predictions, unsupervised clustering for segmentation, dimensionality reduction for analysis, and recommendation systems for personalization. Together with Generative AI, they make up a subset of what AI can do for marketing and creativity.

On Thursday, I'll be discussing with a $400 million fund the areas that may become important in the near future and where the money is likely to flow.

David Beisel, Partner at NextView Ventures, said it best: "the current challenge for founders in the Generative AI category is to build not just a product, but also a defensible business model that can endure. Anyone can put an application skin on these underlying generative engines, but the key is to incorporate sustainable competitive differentiation through strategies like network effects, raised switching costs, ingrained product partnerships, and so on."

I see a number of themes that would fit that description:

4 PREDICTIONS

• Action-driven AI

Also known as agent-based AI, is an exciting area of AI development where models are designed to act as agents that can choose actions on their own. These actions can be simple, like turning on a light or opening a door, or more complex, like scheduling a meeting or trading stocks.

One of the key advantages of action-driven AI is that it allows for the creation of more autonomous systems. For example, in the customer service industry, where AI-powered virtual assistants could handle various tasks, like scheduling appointments, answering frequently asked questions and providing personalized recommendations.

Another area where action-driven AI has potential is in the field of gaming. Game AI agents can become more believable and lifelike if they can act in a way that mimics real-world human behavior. This can enhance the player's immersion and make the game more challenging and engaging.

For further insight, I would suggest exploring the capabilities of Adept. They have a presence on Twitter where they discuss their upcoming offerings in the market, and their website provides a comprehensive overview of their product (link 1) (link 2)

• Emotion and micro-gesture based AI 

Is an area of research that aims to make AI models more human-like by incorporating emotional and physical expressions into their behavior. One example of this is the integration of Synthesia's avatars with new micro-gestures and OpenAI's ChatGPT. This integration allows the avatars to display emotions such as happiness, sadness, and surprise through subtle facial expressions and body language, while also generating natural language responses through ChatGPT. (link)

There's also some concern about the impact of emotion and micro-gesture based AI on society. One major concern is that it could be used to manipulate or deceive people. For example, a virtual assistant could use micro-gestures and emotions to influence users' decisions or make them feel a certain way. Additionally, there's a concern that emotionally expressive AI could be perceived as eerie or uncanny and that it could blur the line between human and machine in ways that make it difficult to trust AI.

The integration of Synthesia's avatars with new micro-gestures + OpenAI's ChatGPT is a first step in this direction and is an exciting development in the field of AI.

• Reinforcement learning

Is a machine learning method that allows models to learn through trial and error. It is useful in situations where the model needs to make decisions based on incomplete information or in an uncertain environment. In the marketing sector, reinforcement learning can be applied to optimize marketing campaigns by training a model to make decisions on which ads to show to which customers, based on real-time data such as browsing and purchase history. Reinforcement learning can also be used to personalize content, by training a model to optimize the content of a website or communication based on the customer's behavior and interests, leading to more engagement, sales, and customer retention.reliability/trust in the completion. Companies like Pega are getting pretty good at this for Fortune 500 clients (in milliseconds response time) but I still haven’t seen anything usable by SME

• Reliability and trust

Important considerations when using AI systems like ChatGPT. While it has been impressive in its ability to generate accurate and deep outputs, it is not perfect and should not be relied upon for tasks that require a high level of precision. However, it can still be a useful tool for improving our own writing, generating text, and making suggestions. The integration of ChatGPT with other platforms like Wolfram|Alpha, for example, means that it could not only produce "plausible human-like output," but also output that is backed by the vast knowledge and computation available through that platform. This integration could be one way to improve the reliability of the models. (link)

So, in short, the best is yet to come… Ladies and Gentlemen, start your engines! (oh wait, did I already mention I used to run TopGear.com for the BBC??)

NEWS

Now Ladies and gentlemen, drumroll please... welcome the newest 1,000 members to our AI party since last week! It's getting crowded in here, but don't worry, I've got you covered. I'm hosting a webinar in January where I'll guide you through the AI-infused marketing funnel, step by step. From top to bottom, we'll cover how AI can help you improve your marketing efforts with demos and a curated list of tools that will knock your socks off. Keep your eyes peeled for more details on the webinar and if you want to secure your spot, pre-book now before it's too late! (link)

TOOLS

And if you're looking for some new tools to add to your marketing toolbox, check out the AI Radar for the full list. (link) Here are a few of my personal favorites from the past week:

-Synthesis X (link) - Generate images without prompts, just like a boss. (link)

-Clip Interrogator - Find the prompt from an image, like a detective. (link)

-Munch (link) - Create thumb-stopping short-form videos, like a pro. (link)

-Prompt extend - Prompt Generator for MidJourney & others, like a superhero. (link)

-Mokker AI - Professional photos of your product, like a supermodel. (link)

LEARN

-This repo is a treasure trove of academic papers, datasets and other goodies on multimodal machine learning (MML) research applied to music. (link)

-Learn how to download those fancy textual inversion embeddings, how to use them, even how to train them yourself with the automatic1111 repository. It's like a choose-your-own-adventure book but for machine learning! (link)

BUILD

-Are you tired of boring old CVs? In this post, you'll learn how to build a CV profile generator that will make your or your candidate resume stand out like a unicorn in a sea of horses. (link)

-And if you're feeling extra adventurous, why not try building something even cooler, like a text2video. So, you've heard of Synthesia type avatar videos but have you heard of text2video? Come on, give it a try. (link)

DID YOU KNOW

-GLM 130B - a new opensource model that's giving GPT-3 175B a run for its money and even giving Google's PALM 540B a good competition. This is gold!!! (link)

-And just when you thought it couldn't get any better, you can even load large models using peer-to-peer like BitTorrent. Mind blown? Me too. Check it out here. (link)

-You thought ChatGPT was just for writing? Think again! People are getting creative and building all sorts of cool apps on top of it. Here are just a few examples that OpenAI has selected for you:

1. Connect your ChatGPT with your Whatsapp for an AI-powered chat experience.(link)

2. ChatGPT Writer: Use ChatGPT to generate emails or replies based on your prompts. (link)

3. WebChatGPT: This web app (link) gives you relevant results from the web like a modern day librarian.

4. YouTube Summary with ChatGPT: Generate text summaries of any YouTube video, perfect for procrastinating.(link)

5. TweetGPT: It's like having a personal AI assistant for your social media game. It uses ChatGPT to write your tweets, reply to comments, and be the life of the party on Twitter. (link)

FINE-TUNE

Brand managers, we can't just copy and paste ChatGPT's completions and call it a day can we? There's a lot more work required to align with our brand values. The solution? Fine-tuning models, and lucky for us, there are some great products out there to simplify the process.

Like Trudo.ai, that works with OpenAI models (link) or Tune My AI (link) for those pictures you generated with stable diffusion

And if you're feeling extra ambitious, you can always learn to do it yourself, because why not? Plus, you can then execute on your own or have a more informed conversation with your favorite agency. This is one of the services offered at https://getinference.com , my agency! Apologies for the subtle plug but someone has to pay for the rising cost of electricity now that 5 million people are using GPUs to fine-tune their models 24/7.

Alright folks, that's all for this week unless there's a breakthrough that needs the spotlight! Oh, one last thing, ever wanted to chat with Bob Marley? (link)