Dangerous Dependencies When Adopting AI for Business
Vendor-driven AI updates can challenge business strategy even after careful testing and best in class integration.
We're still in the early adopter phase of AI. While many companies have adopted AI to please shareholders, excite users, or encourage internal developers, there are still a lot of details to figure out. The well-known challenges are the difficulty of aligning AI tooling with human interests and keeping user data secure. But the bigger challenge for businesses is finding a consistent, reliable, and controllable AI environment to build upon.
This week, DALL·E 3 released an incredible new set of image generation capabilities that are deeply integrated with their revolutionary ChatGPT interface. You can discuss the generated images and give feedback that can add a rich set of details to the images. While we haven't been able to play with the tool directly yet, it's a paradigm shift that's very exciting.
Using a chat client to generate images, users will not need to worry about prompting and can instead rely on the AI model to create prompts that follow their intentions and feedback. This is something no other platform has currently, and it's a big step forward for user experience. ChatGPT has become an art assistant.
This is incredible for users, but what does it look like for businesses? Businesses might look to use DALL·E 3 the way they were looking at Midjourney in the creation of AI artwork. The big benefit of DALL·E is that there will be an API to directly ask for photos, rather than needing to use a Discord client to request images. This makes it way easier to get images and reduces the manual time required to produce and organize them.
But if we look at DALL·E 2, businesses should be wary. DALL·E 2 was updated and changed dramatically over the months after its initial release. Styles and consistent outputs can shift, causing what once worked to produce a branded style to no longer look the same. Imagine if a business's brand design completely changed whenever another company changed their software! This is a big issue for businesses wanting to rely on this tooling for their needs.
ChatGPT similarly has this issue, and most customers are completely unaware of it. This problem gets much more obvious with images. When you've built something cool on top of these tools, you expect it'll keep working the way you built it. For your average user, it isn't such a big deal if things change, but it is an issue for businesses.
This isn't unique to OpenAI, this is a trend in Big Tech. Businesses constantly need to play catch up with Google's updated SEO rules whenever they change. If they don't stay up to date they get fewer customers coming to their site. Google makes these changes to help progress the development of the web in a number of ways that are beneficial for users. But you can't quite ignore that Google makes this happen through threats and algorithmic brute force against businesses. Businesses ultimately pay the bill for these changes.
OpenAI similarly forces businesses to keep up to date. This is in part because of their initiative to align AI with human interests. Older models may be problematic and are regularly trained to be more equitable and socially responsible in their responses.
OpenAI is also focused on research, and that drives these consistent updates. After all, they release these models for a cheap rate so that they can get more user data for training. If they don't keep most users on the latest versions, they'll miss out on valuable data.
While ChatGPT does have versions, those versions eventually get deprecated. Usually only with a few months notice. Whether you're forced to update today or in a few months, the result is still the same. Businesses need to operate on OpenAI's schedule, not their own.
Fortunately, there are a number of open source alternatives being developed and expanded by other companies. Meta is very bold about releasing their latest tooling for businesses to use. This comes from their strategy of directing the market to use their tools, so they can benefit from Open Source contributions and the cheaper talent creation that comes from others outside their company adopting their tooling.
Businesses need to think about how the tooling they use will develop and if they'll get locked into using a high-cost vendor. Imagine paying for electricity, and the electric company requires you to rewire your office every few months. Maybe the electricity gets a little cheaper, but it's not likely to offset the cost for many years. That's the kind of vendor some of these big tech players can be, and businesses need to be wary.
The future holds us all in some balancing act. We need tools that give us more control, flexibility, reliability and consistency to get complex work done. AI gives us a compelling case to solve many of these challenges. But the tools available to companies still have a long way to go before they're ready and you don't want to have to pay the cost for their research and exploration. Reach out to experts familiar with AI before committing to a platform or service.