In my last article I wrote about my sceptisms and the fears and possibilities of AI.
Here I want to define what AI is not. Where are the boundaries? What is our outcome? At the end we defined a small manifest for the use of AI at onexip. It is not an easy task and it will definitely change over time. In the following text, you will find our summary and our outcomes for AI usage.
1. AI is not a super weapon!
Don’t anticipate miracles from any AI. An AI will only provide the outcomes it’s been trained for, and sometimes it might not even manage that, because it is also faulty.
2. AI has not human creativity
If you expect human creativity from an AI, you are wrong. It will only deliver predefined ideas and combinations of it. Expecting entirely new things, is wrong! You will only get things it was trained for and combinations of it. Combining human creativity with AI is the key!
3. AI will not reach superb results
Sure, AI can deliver impressive results, and humans are expected to deliver high-quality results as well. However, an AI can only deliver results it was trained for; it cannot outperform its training data.
4. AI is not the answer for everything
Implementing a trained AI might solve a lot of issues, but there may be easier and more reliable solutions out there. Do not over-utilize AI either as a tool or in your software product.
5. AI cannot lead Humans
This one shall be obvious. A Human will not take orders from an AI seriously. Also, keep in mind that AI cannot take any responsibility for its actions.
6. AI cannot create exceptionally new things
Do not expect AI to raise the bar itself. Exceptional outbreaks will still come from humans in the foreseeable future. Humans are still needed to evaluate results and think out of the box and find new creative solutions.
7. AI is not an A-level designer, A-writer nor A-“software developer”.
The results of an AI may generally be satisfying or good. In school grades something between B and C. An A-grade output is quite rare and from current perspective only achievable in applications where AI is working in a highly learnable environment like feature detection in an image or audio feature detection for Beat Per Minute estimation.
Final Conclusion
Hence, I created our 5 Rules Manifest of AI usage:
1. Do not let AI do your work.
2. AI is a tool, that can help you delivering faster results.
3. Do not use Source Code directly from an AI Output.
4. Always verify the results generated by AI carefully.
5. Think of AI as your brainstorming buddy.
If you ask an AI to generate an image of a “brain surgical device” and you get this image back. You know, you shall re-think it.
About the Author
Matthias
Frequently ask questions:
What is AI not?
AI is not a superweapon. It cannot perform miracles and is fallible. Do not expect human creativity from AI.
What are the limits of AI?
AI cannot create exceptionally new things or replicate human creativity. It only provides predefined ideas and cannot surpass its training data.
Can AI solve all problems?
No, AI is not the answer to everything. There may be simpler and more reliable solutions to some problems.
Can AI lead people?
No, AI cannot lead people.
What are realistic expectations of AI?
AI can serve as a tool to deliver faster results, but its outputs should always be carefully reviewed. Extraordinary breakthroughs will continue to come from humans.
