Artificial intelligence (AI) and machine learning (ML) technologies improve automation and efficiency in data centers and storage infrastructures. Some AI-influenced networks and systems still require heavy human interaction, but recent ventures in intelligence have increased the amount of work that artificial systems and machines can perform, as well as their accuracy. One of these innovations is formative artificial intelligence.
- What Is Formative AI?
- Types of Formative AI
- Generative AI
- Generative Adversarial Networks (GANs)
- Uses for Formative AI
- Potential Drawbacks and Dangers of Formative AI
- AI in Data Centers and Storage
What Is Formative AI?
Formative AI is an emerging artificial intelligence technology that focuses on the ability of AI to generate and process accurate digital content. It first gained widespread attention when Gartner mentioned it in the Gartner Hype Cycle for Emerging Technologies 2020 report. According to recent research, formative AI “dynamically adapts” to changes in systems and data.
Arguably, artificial intelligence is supposed to dynamically change depending on the input of data and the given situation; in this context, formative AI doesn’t sound like anything new. But both AI and machine learning require immense amounts of data input in order to analyze and develop an understanding of that data and how to utilize it. This learning process takes time. Formative AI makes much quicker adjustments based on rapid changes in systems, software, or infrastructure.
Formative AI poses possibilities for data centers and storage infrastructures, as well as networks, threat intelligence systems, and software development. Though it’s in its infancy, formative AI suggests greater intelligence and initiative from computing systems, which ideally will provide more automation and accuracy.
Types of Formative AI
Formative AI can encompass augmented design and development, big data processing, and ontologies and graphs. However, here we’re covering the following two major fields of artificial intelligence in formative systems.
Generative AI is content generation performed by artificially intelligent systems. This can include:
- Photo editing (adding color or making updates where needed)
- Visual art (such as prints) based on existing pieces
- Video editing
- Audio generation, including realistic vocal production and music composition
- Text content and caption generation.
Generative AI is particularly promising for business fields such as content marketing and design — companies that require large, regular amounts of copy and visual marketing content will benefit from its automated design. It can also generate realistic-looking human faces, which don’t belong to any one person but resemble a photograph of a live human being. We’ll come back to this later.
Generative Adversarial Networks (GANs)
A generative adversarial network is a type of machine learning that was originally developed by pitting two neural networks against each other in a game. One neural network is a generative network (similar to the AI discussed above) and the other is a discriminative network. The generator creates synthetic pieces of content, perhaps an image, and the discriminator must identify which content shown to it is generated or real. A discriminatory model receives input variables and must place them in the correct location, while a generative model creates an example.
The benefit of GANs is twofold: developing in networks the ability to generate realistic content (generative) and the ability to distinguish fake, or machine-created content. Each neural network sharpens the other; one improves in creating artificial work, while the other improves in telling the difference. The discriminatory network is important because formative AI, while useful, can be a source of fraud as well as legitimate business content creation.
Uses for Formative AI
New concept though it may be, some major corporations are already implementing or supporting formative AI technologies. These include IBM and NVIDIA. NVIDIA is using generative adversarial networks to train machines to perform predictive analytics from a synthetic set of data; it has also created an AI model that uses a limited pool of data to create synthetic artwork. IBM is researching how generative AI models can increase the discovery phase of drug design through virtual molecule creation.
Synthesis AI, a company that produces synthetic data and trains machines on it, has developed computer vision technology using generative AI. The company’s intent is to shift from heavy human supervised learning to a greater reliance on machines, which can identify images using generative AI and computer vision much more quickly. Camera applications can benefit from GANs, too: these networks learn to identify faces more accurately, which could improve surveillance techniques.
Potential Drawbacks and Dangers of Formative AI
Because generative networks can create realistic content, there is a risk that synthetic images, for example, can be used for fraud. Generative technology could eventually create synthetic video and photo content that falsifies real people. The use of deep learning to falsify people’s words or actions is called a deepfake. Security companies are already pointing out that this false content is a concern.
This isn’t a comprehensive overview of deepfakes or generative AI, but it raises some ethical issues:
- A well-placed, falsified video or photo could ruin someone’s reputation. It could also be used politically, perhaps to influence voters.
- AI has developed so quickly that lawmakers have had little time to keep up, so there aren’t yet many existing legal precedents for handling deepfakes when they come up.
- Generative adversarial networks are designed to both create realistic content and locate synthetic content, so there’s a possibility for AI to be pitted against itself and detect false photos or images, too.
AI in Data Centers and Storage
AI offers a variety of benefits for data centers and storage infrastructures. In addition to the automation of basic administrative and monitoring tasks, which removes burdens from IT professionals, intelligent systems can:
- Shift storage volumes or workloads to locations where they can run most effectively
- Detect problems, such as outages or network slowdowns, more quickly.
- Create reports automatically when such problems arise.
- Help with threat intelligence, alerting IT administration when strange trends arise.
- Perform sensor data analytics once sensor data is uploaded to the cloud from remote locations.
- Update and patch software.
- Dynamically provision servers when they’re needed on short notice.
AI that both adds metadata to data and analyzes existing metadata is particularly important for object storage. In stores and lakes that hold large volumes of unstructured data, artificial intelligence can more efficiently sort or search identifying information about that data. AI and object storage are an important partnership; large organizations need object storage for both its scalability and its lower costs, and they’ll need intelligent automated systems to manage object-stored data.
Data and AI have a somewhat symbiotic relationship in modern data centers and big data applications: in most cases, intelligent systems and machine learning platforms require large volumes of data initially to be trained. But data science also needs AI and machine learning to extract the most relevant insights and implement them to improve storage and infrastructure. AI, predictive analytics, automation, and big data all go hand in hand.
Also Read: Data Storage, AI, and IO Patterns