The company took a whopping 25 million in a Series A funding round led by Greenoaks, and a sizable swag of VC funds. Exafunction is a relative of Google’s machine learning division and is looking to reinvent the wheel of AI training with a subscription based platform. It’s already a winner in its home market of Australia and is catching up in the US. For a start it’s using the latest and greatest hardware and software, such as the GPU and cloud. The company has an army of technical wizards to help out its clients vegamovies.
One of the most interesting features of the platform is that it’s not a black box. The software uses a mix of hardware and virtualization to enable the company’s aforementioned aforementioned aforementioned aforementioned. The best part is that the team is focused on customer centricity, ensuring a top notch experience a la carte. Some of its clients include Microsoft, IBM, IBM Watson, and Google’s DeepMind. As with any new tech, the company has a hefty learning curve but they’re adamant that no one is above the fray. This is a welcome development for an industry that’s often viewed as a tamer to navigate taraftarium24.
The company also demonstrated the impressive feat of demonstrating a notable reduction in energy and resource usage while improving the efficacy of the model a nifty little software platform. One of the bigger prizes is a pilot program aimed at helping out one of the largest insurers in the country with a test and learn program aimed at introducing the latest gizmos to their portfolio.
Institutes for Trustworthy AI
Exafunction is a startup aiming to lower the costs of developing AI by abstracting away hardware. Its platform moves computation onto cheap and efficient hardware, and then dynamically allocates resources to a customer’s application. The company’s executives say they see a huge addressable market. However, they declined to discuss the inner workings of their technology digitalnewshour.
One of the core areas of AI research is inference, which uses machine learning to generalize from data. Core research in the field involves logical reasoning, planning, knowledge representation, and multi-agent systems.
Another area of research involves developing software architectures for AI. This includes automated training of models. ML has made great progress in recent years. Developing state-of-the-art systems requires enormous computing power. New techniques, however, can reduce the requirements.
Creating trustworthy AI systems presents new challenges. These include integrating AI components, incorporating unintended consequences, and working within existing economic structures. To build a successful system, experts must work together.
Institutes for Trustworthy AI will have a key role in developing certification methods and testbeds. They will also collaborate on best practices for building AI systems.
The Institute for AI for Decision Making will focus on bringing decision making principles into AI. It will develop breakthroughs in AI that are framed by these principles. Specifically, its research will focus on knowledge representation, causal modeling, and decision support for real-world use cases.
In addition, it will also collaborate with NIST on evaluation benchmarks and testing methods for artificial intelligence.