After several incidents of Samsung’s confidential documents leaked because of ChatGPT, Samsung has banned its employees from using the ChatGPT model. Samsung isn’t the only company asking employees not to use ChatGPT; other companies, including Apple, are also banning the use of the ChatGPT model in their workspaces and considering disciplinary action against employees who share corporate data with LLMs such as ChatGPT.
To counter this, Samsung is developing its large-scale language model (LLM), Anti-ChatGPT, to prevent sensitive data from being leaked into the ChatGPT model. It’s led by Samsung’s Research division, which started development in early June, investing manpower and resources into the project for internal use.
Samsung’s Anti-ChatGPT to Complete Development Next Month
Samsung aims to release the Alpha version for internal use, with the first LLM version completed in two months. To achieve this, Samsung Electronics has restricted the use of GPUs for all other in-house software development organizations, as the company has invested all its GPU resources into AI development. GPUs are essential for the training of machine learning models.
Samsung has been in talks with various firms, such as OpenAI, Microsoft, Google, and Naver, to collaborate and avoid leaks. This is while keeping employees’ workflows more efficient. The company aims to increase productivity without security concerns and without relying on language models made by OpenAI’s ChatGPT, Microsoft, or Google. However, the company has not decided whether to make its AI solution available to general users.
GPUs are essential for learning human language and data, and Samsung Electronics aims to develop its LLM within two months. The company is the largest foreign investor in Vietnam and has four plants and two R&D centers.
Data accuracy and information are currently among the most significant challenges.
It will focus on employees using it for software development, document summarization, and language translation tasks during its initial release. This is for Samsung’s Device Solutions Division at launch. Samsung believes AI solutions can drastically reduce software development and semiconductor design time.
The company already recognizes the potential risks posed by AI-generated text and aims to protect its proprietary information, intellectual property, and confidential data from leakage or misuse. By developing its own LLM model, the company is exploring ways to block sensitive information uploads and external services. This makes it difficult to retrieve and delete the data, meaning it could be disclosed to other users. The company is concerned that the data transmitted to the AI platform is stored on external servers.
Top-level management meetings decided to develop their AI solution as soon as possible. Various South Korean firms, including Kakao’s KoGPT, LG’s ExaOne, Naver’s HyperCloverX, and SKT’s A-Dot, are developing their LLM-based generative AI solutions to write and draw like humans.
The company uses all available GPU resources for an ultra-large LLM-based AI model.
It is rapidly becoming popular for many companies to work on. Since Samsung licensed AMD’s Radeon GPU IP under a multi-year agreement, it could also be used for ultra-large AI system projects. This development is good news for the AI industry. It will bring more competition and innovation to bring better products for consumers, such as Bard and ChatGPT.
It could threaten Apple and other companies that develop LLMs, as it could be more powerful and efficient than their LLMs. Apple has not been actively talking about artificial intelligence for some time now. However, competition from Samsung could change that, and Apple may have to step up its game and delve deeper into AI.
There isn’t much information available; more is expected in July. However, Samsung may focus too much on AI and neglect other aspects of its business, such as hardware development, like Google, which has shifted its focus to AI, changing the company’s personality.