FOMI - Fear Of Missing AI
Is there a chance of missing out if you don’t jump in and start using modern Artificial Intelligence tools today?
The rapid pace of innovation in the Artificial Intelligence (AI) landscape is impressive and overwhelming, with breakthroughs such as GPT-3 and generative AI art emerging seemingly every day. As businesses and individuals seek to stay ahead of the curve, there is a growing sense of FOMO (Fear Of Missing Out) around AI. However, it's important to remember that not every organization or individual is entirely ready to rely on AI. The speed of changes, the emergence of new capabilities, and societal challenges of adoption all contribute to the difficulty in finding a good foundation for us to fully adopt AI tooling today.
In Summary: In this article, we'll discuss if there is a chance of missing out on AI for individuals and businesses. For many, it's too early to adopt AIs fully. The AI landscape continues to change quickly, bringing new capabilities to leverage. The velocity of change makes the foundation unstable, and many societal challenges may block adoption.
The Speed of Change
Large companies have seen the potential of modern AI techniques and are funding, publishing, and releasing research findings that expand the possibilities of this space. The YouTube channel Two-Minute Papers regularly demonstrates how fast AI techniques are evolving. The open availability of these resources continues to grow today's AI fervor.
This fast change is making cutting-edge techniques from prior years obsolete. We're generating more complicated and accurate content today. It's also becoming cheaper as researchers find efficiencies and improve techniques. At the beginning of 2023, Google researchers announced a new Image Generation AI, Muse, that is ten times faster than OpenAI's DALL·E 2 AI model. Muse takes a very different approach to generating images and could majorly impact the future of Midjourney and DALL·E.
Businesses can build on top of Stable Diffusion or OpenAI's GPT services to create their own offering to customers. However, this dependency brings some risks. Failing to stay up-to-date with the latest AI advances could lead to missed opportunities and a competitive disadvantage. Tying yourself too closely to a short-lived version of a popular model could mean re-building your service multiple times with different AIs to keep away competition.
Outside of businesses, individuals also want to be able to leverage AI. Many AI advocates are looking at the best ways to use this technology which is looking to be a frustrating experience. When you find a good process for creating images or writing prompts, a new AI model releases that performs better and take different inputs to complete the same task. While some prompts may continue to work, many must be completely re-constructed.
Be skeptical of courses that offer "the right way" to use any of these systems. While good advice exists, it's time-sensitive and prone to change. Many prompts for DALL·E 2 have broken over time as OpenAI changed and updated their AI's model for creating images. Relying on large services like DALL·E 2 gives you the latest and greatest performance and quality but at the cost of consistency. It's too early for "Best Practices" for users of these AIs.
The Emergence of New Capabilities
New capabilities also challenge our ability to build a stable and consistent approach to using AI. Returning to Google's new Image Generation Model, Muse, we see the ability to edit photos with just text! This new technique changes how much users rely on existing AI techniques like inpainting to edit images. While inpainting might still be helpful, this descriptive image replacement solves many of the same problems.
Creating a product built on current AI capabilities comes with a risk of disruption. Stable Diffusion opened the floodgates for adopting AI image-generation techniques when they released their code. Now users can create vibrant AI images on their phones using an App! Midjourney is more convenient, faster, and easier to create fantastic art with than running the app on your phone, but Stable Diffusion allows those with time to explore this new world for free. The business model for AI-based companies can change overnight.
Staying abreast of all the new changes in the AI field is a challenging task. Aggregators, like There's An AI for That, provide some means of following the latest changes but with so much activity it's hard to know the latest capabilities. This difficulty is partly why modern AI remains a research focus rather than something ready for larger mainstream and core business adoption.
These challenges don’t necessitate avoiding AI entirely. They warn against novel use cases, rigid workflows, and custom model creation. Capabilities are changing quickly, and tying ourselves too strictly to one way of doing things will lead to problems in the future.
Societal Challenges
One of AI's biggest hurdles is the societal changes necessary for broader adoption. AI is increasing the pressure for existing issues like misinformation which we wrote about in A Positive Side of AI. It's also creating new issues, like the potential for copyright infringement of artists with Image AIs or the potential for libel from generative AIs. Two contradictory arguments, but both of which could dramatically change our perspective of how AIs are trained and used in our society.
One of the largest issues is the ability to copy someone's digital identity using AIs. While there are already many ways to steal someone's identity using a few key pieces of information, more manipulative capabilities are within reach that allow a hacker to perform socially engineering attacks using AIs. It's possible to deepfake images, videos, voice, and style (conversational tone, artistic voice) of a person.
In such an exposed world, we need a safeguard. A digital signature and identification that allows us to encode our interactions to be verified by anyone receiving the associated message. These messages should be able to be published with the encoding included so we have a paper trail for the source of information. North American society needs this already, as this could also help resolve many misinformation concerns by improving attribution. AIs will make it harder to trust digitally sourced communication without such a system.
When we get into the technical details of such a system, there's an overwhelming amount of work to do. Many societal challenges are significant and probably can't be hacked together in a weekend hackathon or by investing in another startup. We need a government or a Google-sized company to take on this challenge.
These societal problems pose a serious issue for anyone looking into AI. For consumers, we can see the fickle nature of AI usage in Lensa losing momentum after a swift climb in new users. Many see the AI images as tacky or unappealing after the quick fad for these images.
For businesses, look no further than the class action lawsuit against image AI companies. Startups are notorious for moving quickly and breaking things but with so many eyes on AI and so much excitement, it may be hard for companies to stay under the radar from litigation.
Leaning into AI has some incredible capabilities for marketing, creating unique experiences, and bringing new value to users. Still, many societal challenges come with adopting these new capabilities. Individuals and businesses need to keep these challenges in mind when leveraging AI.
Moving Ahead without FOMI
As we look over the AI landscape, we can see a lot of potential. Things are changing quickly, new users are flocking to AI solutions, and funding is increasing for further research. AI feels like a technology that is going places when compared to crypto-currency, VR/AR, and the Metaverse. It's natural to feel like you might get left behind if you don't stay up to date.
AIs are here to stay, but we still don't quite grasp where they will fit. But as we've reviewed above, we're still incredibly early. None of the foundations of these technologies have solidified enough to build a stable business that will last a decade or more. Learning techniques is great for better understanding how AIs work but we'll likely need to learn new techniques as new capabilities and methods are developed. Society isn't ready to understand how these AIs will fit into our lives.
We suggest approaching AI with a critical eye while also maintaining an open mind to the possibilities it presents. Staying up-to-date and adopting AI for non-critical tasks and systems will develop some competency without over-reliance, which can get individuals and businesses into trouble. Ultimately, by approaching AI with a balanced and informed perspective, individuals and businesses can position themselves to take advantage of its many benefits while avoiding the potential pitfalls of adopting new and unproven technologies.
 | A guest post by
|