What is generative AI?
What is generative AI?
Generative AI is artificial intelligence (AI) that creates (generates) new content such as text, images, audio, and program code based on instructions from users. A major feature is that they learn a huge amount of data on the Internet in advance and apply that knowledge to produce output as if humans created it. While conventional AI has been good at "analysis" and "prediction" to find specific patterns in data, generative AI is groundbreaking in that it creates content that users have not yet noticed from what it has learned.
The model of Generative AI is called an LLM (Large Language Model), which can generate natural sentences like when you experience when using ChatGPT.
How generative AI works
Generative AI uses machine learning techniques to extract and learn features from vast amounts of data. It is called LLM (Large Language Model). You can use an LLM, for example, in text generation, it is possible to probabilistically predict the next word and create natural sentences. Recently, it has become possible to train LLMs not only on text but also on other digital information such as images (multimodal* compatible) and interpret images.
*Multimodal: Combining multiple types of data, such as images, audio and text.
What to look out for when using generative AI
While expectations for a more convenient and prosperous AI society are rising, concerns about the use of AI are also growing from the perspective of ethics, reliability, and security. It is also important not to take the information generated by AI at face value, but to understand its characteristics and the risks involved before using it.
1. Suspect "bias" or "errors" in the output information
Generative AI responses may not always be accurate or fair. Because it learns from vast amounts of data, biased data can also lead to bias and discriminatory bias in the output. It can also generate false conclusions (hallucinations). It is important to consider AI's answers as "reference information" and for humans to make the final decisions.
2. Ensure security and prevent information leakage
The instructions and questions you input into generative AI are called "prompts." From a security standpoint, it is crucial not to include any sensitive or personal information in the prompt. This is because the input information can be used for AI training data or recorded by service providers, posing a significant risk of unintentional leakage of internal company information, customer data, and personal privacy.
To ensure complete security, it is necessary to always check the terms of use before using the service and to strictly enforce the rule that sensitive information should never be entered as a prompt.
3. Fulfill legal and social responsibilities based on AI ethics
AI ethics strongly urges those of us who use generative AI to fulfill our legal and social responsibilities, including copyright.
- Respect for others' rights: From an AI ethics perspective, respecting the intellectual property rights of others is a fundamental principle. The potential for AI-generated works to closely resemble existing copyrighted works does not only incur a legal risk of copyright infringement but also an ethical issue of disrespecting the rights of creators. Especially when using it commercially, it is essential to carefully verify that the product does not infringe on the rights of others.
- Compliance with laws and social norms: AI ethics also requires adapting to the ever-changing rules of society. AI laws and regulations are in the process of being developed around the world, and there is a risk of unknowingly violating the law. From an ethical perspective of fulfilling our responsibility to society, it is important to always pay attention to the latest trends and consult with experts if necessary.
Benefits of using generative AI in companies
With the advent of conversational generative AI such as ChatGPT, expectations for generative AI in improving business efficiency are increasing. These AIs can revolutionize a wide range of routine tasks, such as creating reports and meeting minutes, summarizing materials, generating and verifying programming code, and responding to customer inquiries.
-
Significantly improve operational efficiency and reduce costs
By replacing routine documentation, data analysis, and information gathering, employees can focus on more strategic and creative tasks. This can lead to a reduction in operating costs, including labor costs, and an increase in overall productivity. -
Faster decision-making and improved quality
In R&D and analysis work, it is possible to quickly extract the necessary information from a huge amount of data and present analysis results from multiple perspectives to accelerate decision-making and improve the accuracy of decision-making by management and personnel. -
Creating new business value
Generative AI has the potential to create new business opportunities that will increase a company's competitiveness, such as developing new services, formulating marketing strategies, and providing personalized customer experiences, due to its ability to generate unprecedented ideas and content.
However, to reap these benefits, it is necessary to overcome the challenges faced by many companies, such as unclear specific utilization methods, risk of information leakage, and implementation and operation costs. In particular, the scarcity of human resources with specialized knowledge of AI technology and the lack of an environment where generative AI can be safe and easily tested are often barriers to adoption.
What is the generative AI that Fujitsu is working on?
Fujitsu provides highly reliable and secure AI and generative AI that expands the productivity and creativity of its customers through Fujitsu Kozuchi, an AI platform that includes cutting-edge technologies from Fujitsu Laboratories.
Fujitsu Kozuchi's features include using the Enterprise Generative AI Framework to enhance the reliability of generative AI, providing the best dedicated model tailored to corporate needs through generative AI reconstruction technology, and using Knowledge Graph Extended RAG to structure and handle corporate information.
We provide reliable and highly accurate output. In addition, we will accelerate the introduction of AI by our customers with our overwhelming "technological capabilities" that fuse cutting-edge computing technology represented by the supercomputer "Fugaku". With a wealth of experience cultivated over more than 30 years of implementation, we have a deep understanding of our customers' operations, so we can propose truly valuable AI utilizations and support the realization of transformation.
As an environment for students to try out Fujitsu's advanced technologies as soon as possible, APIs and web applications for technical components are available free of charge on the Fujitsu Research Portal.
By combining these unique technological capabilities, world-class computing capabilities, and extensive achievements, Fujitsu is promoting the social implementation of generative AI not only by providing technology, but also as a reliable partner that is close to our customers' businesses.
Frequently asked questions
Q. What is the difference between traditional AI and generative AI?
A. Conventional AI is a technology that feeds data to learn and makes optimal judgments, classifications, and predictions. For example, conventional AI is good at analyzing patterns based on known data, such as recognizing objects from images or predicting future sales from past data and deriving fixed answers.
On the other hand, generative AI is a technology that creates new content (text, images, audio, video, etc.) based on the vast amount of data it has learned.
In addition, conventional AI requires experts who understand AI technology when using it. Generative AI, on the other hand, is very different in that it can be used even if you are not an AI expert.
Q. What are the different types of generative AI?
A. There are several types of generative AI. The main ones include text generation, image generation, voice generation, video generation, code generation, etc.
Q. What are the features of Fujitsu's generative AI?
A. Fujitsu's generative AI is characterized by providing the latest enterprise-specific AI technology tailored to corporate needs.
In other words, by using the enterprise generative AI framework to enhance the reliability of generative AI, we can provide the best dedicated model tailored to corporate needs with generative AI reconstruction technology, and by using knowledge graph extension RAG to structure and handle corporate information. It is characterized by providing reliable and high-precision output.
