English | EN 
Navigating Tomorrow – Forward Focus
Gen AI will continue to embed itself into many aspects of our lives as we move slowly towards the possible ultimate expression of AI itself – the ability for a machine to think and act on its own accord.
Outlook #1 – Goodbye hallucinations, hello Prompt Engineering
In the early days of Gen AI, hallucinations were one of the main points of discussion. An evocative phrase, it described the errors which AI models are prone to generating. However, I think that hallucinations are in part due to a fundamental misunderstanding of Large Language Models (LLMs): just because an AI model has been trained on a vast quantity of data to learn a language, it does not mean that it is a Large Knowledge Model.
With this in mind, from now on, I think that the power of prompt engineering and the use of Retrieval-Augmented Generation or RAGs will become widespread – and known to the masses.
RAGs are the documents or data which you give to an LLM to become an ‘expert’ in, and by using RAGs, you are creating a bespoke LLM solution which has expertise in an area.
Outlook #2 – More with same versus same with less
One of the main concerns people have with AI is its influence on the workplace and in particular, on jobs. While AI provides great value to businesses by automating manual tasks, will it also drive unemployment by laying off people?
Throughout history, there have been many examples of the introduction of technology which led to more being done, but with less. And in all times when new technology offers this opportunity, employment goes up. This has been the case from the steam machine in early industrialization, tractors in agriculture, and robots in manufacturing to today’s digitalization.
AI will provide, not only the next level of automation but also the next level of employment. The businesses which will thrive will be those which embrace AI’s productivity benefits; by doing so, they will add to their existing workforce, and not reduce it. Of course, AI will shake up existing roles, and people will have to learn new ways of working. We will learn to work alongside AI in jobs which are being created by a fusion of old and new.
Outlook #3 – Forget about numbers
In the early days of Gen AI, we were focused on numbers. In particular, with BERT, GPT-2, mT5 and GPT-3 it made sense to count the billions and trillions of parameters in the transformer models.
By way of an example, I would be surprised if you still count the number of hertz in your CPU. I remember my first computer; it was an Intel 8088 with 8Mhz frequency. But today, I don’t know what CPU frequency my laptop has, and besides, what is the point as multiple cores are running in parallel with new types of calculation capabilities.
We will quickly see the same development in GPT models. Models will become larger, but other models will become more widespread, such as Small Language Models (SLMs) which already exist. LLMs and SLMs will work together, running tasks in parallel or in sequence. In fact, smaller ones will be better in many aspects. Computing costs will go down, and with that sustainability becomes better. Lastly, models will be built for a specific purpose, with your data and use case in mind, do they really need to be that big.
Outlook #4 – Next level of Gen AI: ReAct
There have been two momentous days in the history of Gen AI.
On 30 November 2022, Sam Altman and OpenAI released ChatGPT. AI, as we know it today, was shaped with ChatGPT and the new world where AI assists us in our daily work has arrived. Machines can now be creative, depending on how you define creativity, of course, and the result is that AI can increase productivity.
But is that it? For me, the answer is no, it’s just the beginning.
Let’s look at the second momentous day: 6 November 2023. Sam Altman was on stage again and the venue was OpenAI’s first developer conference in San Francisco. He was presenting Reasoning and Actioning, packaged as GPTs, and in my view, it is bigger than ChatGPT itself. The research on the field was roughly a year old, and OpenAI managed to turn it into a fully working solution which will eventually change digital life as we know it.
With just your voice giving an instruction, ChatGPT will reason to find the right path and then perform an action based on its reasoning. Of course, with the right level of Human-in-the-Loop according to your preferences.
This is the next level for ChatGPT and brings to mind futuristic movies with talking AI assistants steering spaceships while you sit back reading a book. And while we need to wait a while for the spaceship, we’re very close to having a reasoning AI assistant.
Outlook #5 – Trustworthy AI
On March 13th, the European parliament approved the Artificial Intelligence Act, which ensures safety and compliance with fundamental rights. By this, the Ethical guidelines which has been leading the way for AI implementations is now enforced by law.
Businesses which implement any AI solution, may it be Generative AI or not, will from now on gradually implement layers of Quality and Trust into their solutions. By doing so we will not only comply with the new regulations, but also speed up adoption. Quality and Trust is two crucial cornerstones needed for any user to embrace a new solution, in particular AI solutions.
Finally, there is a guide to how, and to what level depending on use case, we need to address the different ethical perspectives. Quality in AI and Gen AI solutions will become a big new topic, and I foresee it will help tear down the resistance to adoption.
Outlook #5 – 2024: a year without POCs and Pilots
And finally, the following is not an insight into how Gen AI will fare over the coming year but more of a wish list for 2024 and beyond. When it comes to AI, it’s time to end pilots and proof of concepts (POCs). Simply put, we all know that Gen AI works. It has proven itself to work in many ways, and with Gen AI we can all see for ourselves the power that lies in automation. Yet we are so obsessed with proving it again and again in countless POCs and pilots.
Let the coming year be about transformation and adoption. By bringing data together in the right foundation with stewardship to ensure quality, the focus will become the transformation of your business.
If we’re going to try things out, let’s do it in POVs instead. Proof of Value for the business is crucial for adoption and setting the stage for transformation. By bringing in the end-user from the start to understand the new tool and how it will change business processes, transformation will follow.
