A revolution through artificial intelligence (AI) is brewing. ChatGPT is the symptom of this: between the progress it promises and the risks it poses, professionals in all sectors seek to seize the opportunities offered by the tool as much as to guard against its risks. Back to the implications of ChatGPT for the different professions of economic intelligence (IE).
Despite its $10 million investment by Microsoft and an estimated value of $29 billion in 2023, the Open AI LP company has only recently become the focus of attention. In question, the launch of ChatGPT, welcomed by a rush of users, whose number reached 100 million in January.
ChatGPT is a conversational tool, i.e. able to maintain a human-like conversation – like the widely used trading bots. Its particularity lies in the AI on which it is based: version 3.5 of “Generative Pre-trained Transformer” (GPT), a generative AI capable of a form of creation and prediction. Version after version, GPT has seen its number of parameters increase to reach today 175 billion parameters – i.e. fragments of knowledge, taken from online sources such as Common Crawl, Wikipedia and various literary corpora. These parameters constitute the data set on which the learning system is trained. Its Transformer architecture then allows it to answer complex questions by simultaneously processing many elements. Finally, reinforcement learning (RLHF) provides GPT with a complementary training dataset, based on feedback from humans on ChatGPT proposals.
ChatGPT could be particularly useful for open source research. First of all, it has a major advantage, the tool indeed masters nearly a hundred natural languages, thus facilitating access to foreign information. Thanks to its semantic analysis capacity, ChatGPT also makes it possible to identify keywords and resources on a given monitoring theme, a fundamental step that has not yet been automated. It could also make it possible to automate the tasks of labeling the content brought up by the day before. It also has the ability to manage very large amounts of data: by providing it with qualified and up-to-date information, ChatGPT could write daily monitoring, or even carry out due diligence.
From this mass of data, ChatGPT is also able to draw trends. By training it on sensitive data such as intelligence service reports, GPT could detect weak signals and erroneous information. IARPA is also developing a similar AI, the REASON project, to provide new lines of thought for analysts.
However, several limitations should be noted: the GPT database is currently limited to content dating from 2021, which impacts the freshness of its responses. This obstacle could be overcome by connecting the tool to the web and thus providing real-time information. Also, the relevance of the information provided raises questions: ChatGPT sometimes gives erroneous information, which is moreover linked to its intrinsic functioning because the tool only predicts the next word, the probability thus taking precedence over the veracity and the quality of information.
Another problem lies in the sourcing of information, because the databases used by Open AI are relatively opaque. Moreover, ChatGPT does not provide the sources of the information it provides, which requires tedious verification.
If it is capable of carrying out a number of tasks related to the search for information, this tool would benefit from being coupled with other intelligence methods (HUMINT, SIGINT, etc.) to become a real analysis support tool, and ultimately to the decision.
In terms of analysis, ChatGPT struggles to offer reliable avenues for analysis and forecasting, due to the limitations mentioned above – lack of freshness of its database and lack of sourcing. However, it is a very effective synthesis tool if it is trained on the appropriate data, it is thus able to make listings or build managerial analysis grids. It can, for example, provide SWOT analyses, a PESTEL matrix or even assess Porter’s forces in a market.
It therefore appears to be a valuable tool for strategic analysis: to study a market, provide analysis indicators and facilitate the rapid and synthetic visualization of sometimes complex environments. Regarding sector analysis, ChatGPT is also able to provide relevant recommendations in terms of influence, for example by mapping entities, stakeholders and their positions.
Rather than a stand-alone tool that can produce analyzes independently, ChatGPT acts as a facilitator of the strategic analysis that underpins any decision.
The AI revolution is brewing and ChatGPT is only the visible part of it. Despite the many decision-support opportunities it offers, considerable limitations and biases exist. From these stems several recommendations.
From the point of view of the analyst or the watcher, the limits of ChatGPT can be mitigated by cross-referencing the results, from alternative methods or from several AI tools.
From the more general point of view of IE players, it seems judicious that the imminent launch of GPT 4 should be monitored, since the number of its parameters could be multiplied by 500. It is also advisable to acquire suitable internal capacities to master these tools for the benefit of IE activities, particularly in terms of human resources in “prompt engineering”.
Finally, it is advisable to remain vigilant as to the legal vagueness surrounding ChatGPT – which the existing regulations (DSA, IA Act, etc.) struggle to cover – and the reputational risk that a user unaware of his limits could run.
Continue Reading On Medium: