Jacob Posel is a software engineer at e-commerce agency Common Thread Collective, where he focuses on strategies to integrate artificial intelligence into businesses. He says the best use of artificial intelligence is to streamline business processes that would otherwise rely on virtual assistants and cheaper labor.
In a recent conversation, he spoke about AI and human creativity, image generation, costs, and more. The full audio of that discussion is embedded below. The transcript has been edited for length and clarity.
Eric Bandholz: Please briefly tell us about your job.
Jacob Posel: I’m a Senior Software Engineer at Common Thread Collective. I spend most of my time integrating artificial intelligence into creative and commercial processes, most recently in image generation, and my work spans the entire Creative Operating System.
The best use cases for AI are everyday business processes, especially those that are assigned to virtual assistants or alternative forms of work. These tasks are typically well suited for AI. However, since AI has become a buzzword, I would like to define what I think AI is today.
Most people think of AI as large language models, but AI is broader than that. For business processes, it’s systems that have a strong enough view of the world to understand human conversation and text and develop intuition. Start by analyzing those processes to determine how you and your team can be more efficient. What tools do you have? Once you’ve determined the process and how it fits into your business, how can you fully automate it?
Eric Bandholz: Can you give an example?
Jacob Posel: This helps you get a better overall picture of your business. For example, you can bring in sales data or reviews. You can bring that into a Google spreadsheet if you like, and come up with the insights you want to get from that data and the action items that follow. Then you explain that to the LLM and the AI. You share the data you brought in and explain your thought process. The AI can then summarize the data, provide insights, or let you know if there’s anything worth paying attention to.
Eric Bandholz: How do we preserve the core skills of human creativity?
Jacob Posel: I read a research paper on this where they trained an LLM or AI model on its output to see how many iterations it took before the whole process failed. After about 10 iterations it spit out gibberish. If you think about it, 80% of the code on the internet is written by AI, as is much of the online text. So there is a real concern that we will run out of training data to develop new models with and eventually reach a point where these new models can’t make any more progress.
Models are trying to scrape YouTube,videos to get more information. But a lot of very smart people are coming up with,different techniques to improve these models, not just the training data.,Currently, most models just grab as much training data as,possibly spending as much money on computation as possible,and see the results.,This cannot continue indefinitely.
The overall point is that A.I. Empowering People Build your own software. At this point you can build anything you want. Even if you’re not tech savvy, investing some time into setting up the best tech might be frustrating, difficult, or imperfect at first, but you can still do it. The Future of Programming The language will not be Python, JavaScript, or SQL. The next iteration will be natural language. At this point, I think that’s pretty certain.
Eric Bandholz: You use AI to generate the images. How do you do that?
Jacob Posel: The foundational model I’ve been using so far is called Flux, which is different from the Midjourney model. You can tweak your own models. I mainly use Replicate, which is an interface that allows you to interact with the Graphics Processing Unit and tweak your own models.
Mid-journey is amazing Generate an image Based on the text you provided. If I wanted to create an image of a random guy sitting in an armchair under a tree by a lake, I would use Midjourney. But Create an image To handle specific things like real-world products or people, you need to train your own custom models, which Midjourney can’t do, which is why I use Flux.
One thing to keep in mind is that the more specific you make your product, the less creativity your model can provide in the background and other parts of the image. In other words, a very simple product like a t-shirt can be worn by anyone anywhere, but if it needs to be very specific, the model will be overly focused on the product and will have a hard time doing the rest of the image justice.
Training data is very important: if you need a specific angle, provide photos from that specific angle (preferably multiple times) and make sure they are high resolution.
Eric Bandholz: What is the cost?
Jacob Posel: Video is the most expensive so far. The cost is obviously determined by the text, images, and video. For example, Runway uses a credit system. It’s calculated in dollars per credit. The unlimited plan isn’t too bad. It’s around $100 per month. It’s not the cheapest plan in the world, but it’s not exorbitant. It’s more costly in terms of time, and it takes time to master the prompts.
Text to image conversion is a bit more complicated to explain more clearly. Text to Video It tells you how many images are combined. It gets more expensive and more difficult to do right. You have to develop a sense for the language they use to train these models. The more advanced you get, the better you can understand the language of photography and film. But that’s why more advanced tools get more complex and expensive.
The best thing to do is roll up your sleeves and figure it out yourself. The best way to learn Because AI has a personality at this point and you can’t learn everything by just reading about it. That’s the way I think about AI. You need to understand what it responds to and how to get it to do what you want it to do.
Start thinking about your business as different systems and processes. Don’t think of creating an ad as one thing. Break it down into its core steps and keep that perspective and foundation in mind. That’s how you build an engineering product. And that’s where AI comes in. It’s also important to communicate with people who understand AI and how it integrates into your business.
Eric Bandholz: Where can we follow you?
