gliff.ai was founded on the belief that imaging Artificial Intelligence (AI) can make the world a better place.
We are committed to advancing the state-of-the-art whilst practising and encouraging what we call FATEful AI.
F – fair, all AIs should treat people fairly. This means accounting for bias in datasets, AI models and humans.
A – accountable, people are accountable for the AI systems they develop, implement and use. We believe that this requires keeping domain experts (such as clinicians, engineers or scientists), and not just AI experts, in the loop for the whole AI development lifecycle.
T – transparent, AI should be developed with human understanding in mind. We believe that developing AI with a strong audit trail helps users, developers, domain experts and regulatory bodies understand both how an AI has been developed and why it may behave in a particular way. As gliff.ai grows we will also be developing tools that help users interpret and understand their AI models.
E – ethical, AI should be developed and deployed in an ethical fashion. This includes maintaining the privacy and security of individuals’ data and AI systems being used in a reliable and safe manner.
gliff.ai believes that FATEful AI leads to better AI and trusted AI.
How does gliff.ai support FATEful AI?
gliff.ai is developing “MLOps” products – products that support others in developing AI systems. So how does gliff.ai embed FATEful AI practices in their products?
F – how developers make an AI fair comes largely down to knowledge of the domain, the datasets and models being used; gliff.ai encourages fair AI by ensuring that domain experts, data experts and regulatory experts can be involved throughout the development of an AI system.
A – in order to aid developers, deployers and regulators in understanding accountability all gliff.ai products provide a human-readable audit of all actions taken on datasets, such as annotations or data cleaning, and AI models, such as training and testing.
T – although we aren’t developing AI models ourselves, we still take the concept of transparency seriously by making our core software available with open source software licenses; we also use other open source projects to help us build our software and actively encourage our team to contribute back.
E – at gliff.ai we’re trying to do MLOps differently – with privacy and security at the heart of our products by using end-to-end encryption to ensure that nobody sees your data or AI models without your knowledge and explicit permission.
If you would like to know more about how gliff.ai’s ethical approach can help you or your team change the world with AI then please contact us.