The latest in AI research
Artificial Intelligence (AI) is one of the fastest growing technologies in the world and ChatGPT is one of the most advanced models on the market. ChatGPT (for “Chat Generative Pre-trained Transformer”) is a powerful tool that can help solve many complex problems in the world. It offers us innumerable possibilities, including simple design solutions, but also complex applications in medicine, such as regarding cancer detection, and other areas of science and technology.
What is ChatGPT?
ChatGPT is basically a chatbot first of all – the special feature results from the AI behind it, which is based on a language model system. This language model system features deep-learning aspects focused on allowing a human-like conversational flow to emerge.
The model is trained to generate full code or text of any length and style from a prompt (e.g., a question) based on analysis of large amounts of text and pattern recognition. For example, the model may be asked to list creative suggestions for research niches, write a full business plan, or formulate an acceptance speech.
In particular, the GPT-4 versioning is characterized by a high degree of multimodality, which means nothing else than: ChatGPT offers the possibility to accept or generate inputs and outputs of different kinds, which includes text in different forms as well as images. This allows ChatGPT to work with input data from Excel sheets, images, text documents and more.
Overall, ChatGPT has the potential to change our world – if we let it.
ChatGPT as a paradigm shift: between opportunities and risks
The innovative power of ChatGPT is mainly due to its ability to mimic human communication – since communication is an essential factor of human connection and contact, it is not infrequently equated with a certain form of intelligence. The innovative power of ChatGPT is mainly due to its ability to mimic human communication – since communication is an essential factor of human connection and contact, it is not infrequently equated with a certain form of intelligence.
However, even besides this aspect, ChatGPT has quite practical features that can be used for different areas: It has huge capacities in terms of handling data, can analyze large amounts of text and numbers, recognize correlations and patterns on this basis to generate new insights. As a result, it impresses with a wide range of application possibilities, such as the automatic creation of texts for marketing campaigns or the automation of customer support.
This diversity in terms of application options is already clearly reflected in the new offerings that have emerged, as countless new online services are already available within the first few weeks of the release. Likewise, user guides are published that deal with how to improve the communication with ChatGPT in order to get better results.
Daraus ergeben sich jedoch zugleich auch Risiken hinsichtlich der Arbeitsplatzsicherheit, denn wenn ein Großteil der insbesondere kreativen oder analytischen Arbeit nun durch die KI übernommen wird, bedeutet dies in einem Arbeitskontext auch, dass Stundenlöhne sowie Arbeitsanforderungen neu gedacht werden müssen. At the same time, however, this also gives rise to risks with regard to job security, because if a large part of the work, particularly creative or analytical work, is now taken over by AI, this also means in a labor context that hourly wages as well as work requirements have to be rethought. Similar to the introduction of calculators and the resulting protests of mathematics teachers, the goal here is not to enact a ban, but to integrate the use of the application into teaching in a valuable way and to make use of it.
Furthermore, a very practical risk also results from energy consumption, as the ongoing power consumption for each question to ChatGPT was estimated to be up to 1 at the end of March 2023. Thus, a smartphone can be charged up to 60 times with each response from the chatbot. At that time, ChatGPT was already incurring estimated costs of $100,000 to $700,000 per day to operate.
Forecast and critical classification
As a data science company, we deal with the challenge of how to handle data on a daily basis: Even without ChatGPT, questions have arisen here about how data can be analyzed without bias, how evaluations can be made comprehensible in a way that is accessible to the broad masses, and how data can generally be collected and analyzed without incorporating one’s own emotions and opinions.
ChatGPT raises further ethical questions here, which should primarily also address how to evaluate the work with ChatGPT. Especially in the university context it seems so easy to type your requirements into the input field, collect the results and send them nicely formatted in a Word document to the lecturers. However, users of these features should be aware that a large part of the sources that ChatGPT indicates do not exist. The fact that the language model is always convinced of itself is also a risk, because here the user has to check especially meticulously whether correlations between theories are correct.
As users ourselves, we must ensure that our interactions are ethical and responsible, and that we retain control over all decisions. Likewise, the language model must be constantly reviewed: Ultimately, it is a machine that does not think to the same extent as we do – but if we work with its results here, we must ensure that they are correct before we continue to use them.
Overall, ChatGPT has the potential to change our world – if we let it. ChatGPT can be used to redesign areas of life and work and to rethink work processes. If this is done thoughtfully and responsibly, not even a hundredth of the possibilities that arise from the use of AI have been exhausted.
Dr. Kay Stankov
Head of Data Science, Ainovate GmbH