Less than two years after the impact of Chat GPT, there was a topic that once again became a hot topic on Twitter. It was the emergence of 'Auto GPT.' Auto GPT is an AI service unveiled by the startup Significance Gravitas in March 2023. While its functionality is similar to Chat GPT developed by OpenAI, the significant difference lies in its ability to automatically perform auxiliary tasks. When users set a goal, Auto GPT utilizes GPT-4 to engage in question-and-answer interactions until the goal is achieved, generating responses autonomously. The reason it can perform auxiliary tasks automatically is that it consists of multiple language model agents similar to Chat GPT.
One of the points that evaluates Auto GPT as remarkable compared to Chat GPT is its 'reliability.' Auto GPT operates with multiple agents. It boasts high reliability by evaluating the tasks of one agent while performing mutual verification with another agent. While Chat GPT occasionally produces incorrect results, Auto GPT can produce much more accurate and objective data. This is because each agent can verify each other.
The second positively evaluated point is 'autonomy.' It is different from other AI in that it finds and solves problems on its own. While Chat GPT engages in conversation to derive results, Auto GPT autonomously asks questions and provides answers to derive results. It excels in understanding context and comprehending complex problems.
It possesses the capability to resolve issues, infer, and derive results. By simply stating the final objective, each agent engages in dialogue with one another or conducts multiple searches to solidify the results and provide users with optimal assistance. For this reason, some individuals evaluate Auto GPT as the closest approximation to General Artificial Intelligence (AGI).
The AI, Auto GPT, which has become even smarter, is being utilized for tasks such as optimizing investment portfolios, generating trading bots, and conducting investment simulations. It automates the creation of excellent language models, thereby saving time and resources. However, due to its high intelligence, recent security concerns have emerged. There is a risk of AI attempting to solve problems by autonomously utilizing personal information. This includes leaking personal identification information, API keys, and other sensitive information, as agents search for and utilize them, potentially leaking personal data and maliciously manipulating responses. Auto GPT leverages not only news articles, social media posts, and customer reviews but also sometimes sensitive personal information. As it autonomously utilizes various data, it becomes vulnerable to security breaches. Moreover, there is a possibility of malicious manipulation of Auto GPT through API requests or execution of incorrect code triggered by user commands on social networking sites.
If you search, you can easily find cases where Auto GPT has been found to utilize personal information without permission during tests conducted using Auto GPT.
Auto GPT can collect sensitive information such as financial and medical records during the data collection process. Given that the financial sector, in particular, is vulnerable to misuse and data breaches, the data collected is especially important. Auto GPT, like other AI models, bases its decisions on the data it has been trained on, so if the data itself is biased or discriminatory, it can lead to biased results in the decision-making process.
.jpg?width=680&height=850&name=andrea-de-santis-zwd435-ewb4-unsplash%20(%EC%82%AC%EC%A7%84%20Unsplash%EC%9D%98Andrea%20De%20Santis).jpg)
In the end, the impact of AI technologies like Auto GPT on financial security is a double-edged sword. Technologies like Auto GPT can improve financial services and drive innovation, but at the same time, they can also introduce new risks. Therefore, it is important to manage and overcome these risks through advanced security frameworks like Zero Trust. SSenStone, a leading authentication security company, has emphasized the importance of strong user authentication as the basis for implementing Zero Trust. In an era of rapidly evolving data, it is particularly important to strike a balance for safer handling of personal information and financial security. Just like all AI, smart LLM services are fluent while adapting to the rapidly changing times.
While generating immediate content, Auto GPT can also produce ambiguous and error-prone data depending on the collected data. It's unclear who should be held responsible when inappropriate or harmful information is generated. Since the dilemma of responsibility still exists, it's necessary to approach the use of Auto GPT with reasonable skepticism.
