Let's Talk about AI - About the Hype, Experts, Risks and Potentials
America innovates, Asia replicates, Europe regulates.
Currently I see a lot of discussion about the potential of AI and the right way to leverage it. We Germans in particular would be too conservative here and would not utilise the opportunities offered by AI.
AI should already be taught to children at school. (LinkedIn, german)
AI is not an (IT) project (LinkedIn, german)
Companies just need the right AI strategy. (LinkedIn, german, rather from comments)
Germany must dare and invest more. (LinkedIn, german)
GenAI brings value rather for workers than for the company (LinkedIn)
…
The AI Expert Bubble
If it comes to AI, we have a phenomenon like the football World Cup, where everyone knows better than the national coach. Or during COVID, when everyone was a doctor and virus expert.
I just learned that there is a “AI expert” bubble of people with a backround in sales, coaching, or education. Many of them just did a AI training in the last months. They ignore classical AI (Machine Learning) and Data Science, means AI = GenAI for them.
Not sure what I should think, but be sure this flood of “AI experts” is a sign of the hype and will burn money and trust in the next months.
Where AI Comes From
AI has come a long way and is here to stay. The term AI was coined 1956. And ChatGPT is not AI, it is a chat app based on a LLM (GPT4).
Rather more people think, everything started out of the blue. But the paper behind LLMs is from 2017. Generative Adversarial Networks are from 2014. LLMs are a specific area within Natural Language Processing (NLP) which is there since the 1940s and today considered a a part of AI besides Computer Vision, Machine Learning, Robotics, …
It is true, a model like GPT-4o work in areas beyond language. Is this efficient? Mostly not! Is this cool? Yes! Is this AGI (Artificial General Intelligence)? Who cares?! Machines/computers ARE already faster and more intelligent then humens are. At least in many cases. But maybe we have just seen to many films…
Is AI Dangerous?
Can AI be dangerous? In ways you can't imagine. Like a kitchen knive. It depends on us how we use it.
Due to OpenAI’s GPT-4o System Card they had to mitigate the following risks:
The EU AI Act describ the following and more Prohibited AI Practices in Article 5. It shall be prohibited to place on the market or put into service or the use of an:
AI system that deploys subliminal techniques beyond a person’s consciousness or purposefully manipulative or deceptive techniques, …
AI system that exploits any of the vulnerabilities of a natural person or a specific group of persons …
AI systems to infer emotions of a natural person in the areas of workplace and education institutions…
…
Furthermore you will also have some additional insights reading these awful AI cases.
Is AI Overhyped?
Let's come down to earth, if we talk about Generative AI.
GenAI is currently overhyped!
“Classical” AI is currently neglected and investments are low because everyone invest in the fancy GenAI.
Nvidea is earning a lot of money for GPUs and Microsoft for offering GenAI services. All end users are experimenting, but most are not really earning money.
GenAI is for most companies a better ChatBot after the disappointment of the former Chatbot wave. And be honest, ChatGPT is better than yours. (interesting LinkedIn read)
The productivity case for GenAI are Co-Pilots. For coding they are possibly producing hallucinations and mediocre code. Ok for beginners and as assistent it works often well if you know what you are doing. For sure also for learning new things in all areas. (interesting discussion on X)
I see BI and analytics vendors integrating GenAI. But NLP support is not new and finally to talk with your BI is often not really efficient.
Don't get me wrong, GenAI is cool and indeed it is helpful. The potential is real but you have to see beyond the hype.
If you want to understand more I recommend reading the paper “A Business Intelligence System” from 1958 written by Hans Peter Luhn. The ideas are not new.
Who is Driving and Leading AI
I was just involved in a discussion, that “AI” has to be led by the CEO and is no IT topic. So all these so called “AI experts” confirmed that with high conviction. Most of the enterprises I know have to bring digitalization forward first (maybe a typical german thinking) and build structures, skills and process to improve their digital maturity. If you already have this, work on classic AI cases and try out GenAI, but don't bet your business on it.
If we have a look on the latest major events in Data & Analytics like Snowflake Summit, SAP Sapphire, Databricks Data + AI Summit - we are currently inundated with news about "my AI is better than yours". Lots of marketing. Mainly around GenAI. It's hard to distinguish what's really an advance in the market.
The potential of AI is huge and yes, possibly it will transform the world we are living in. Amaras Law tells us "We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run”. Often we want to get the quick win without doing our homework.
As I have shown in my former article “Let’s Talk about AI” you have to consider different worlds, where and how AI can be used. It is a broad field with a huge potential and is coming in many forms.
If you want to enhance you operational world, look for Process Mining and Robotic Process Automation (RPA). Also check the roadmap of your vendors as they typically enhance operational systems more and more with AI.
If you have good data and real potential using forecasting, predictive analytics and pattern detection, build up a data & analytics plattform as your ai backbone to support operations and a broad usage of AI.
If you want to support your employees, watch for AI-Assistents, Augmented Analytics and chatbots based on LLMs using Retrival Augmented Generation (RAG). For the germans, have a look here.
But also do your homework, invest in Data & AI governance, DataOps processes and data quality. Build your data platform for current and future use cases. Enable your data team and users by establishing the right data culture.
Conclusion
Maybe it is right and “America innovates, Asia replicates, Europe regulates”. I currently see a huge benefit in the EU AI Act and we are not the only one fostering regulations. The EU AI Act makes us think about AI. It demands for AI literacy and set boarders where needed. Even before we have seen AI ethics and responsible AI coming up. As a modern society, we must also take responsibility. This should not detract from our innovative strength and the realisation of potential. And be aware of the wave of upcoming “AI experts”. The basics haven’t changed, your IT department is more important than ever and you have to put in the work.