When Southern Cross Austereo (SCA) become an early-stage investor in Melbourne-based Sonnant, Which-50.com decided to reach out to the company and ask what SCA was getting for their money.

Surprisingly, Sonnant CEO Tony Simmons responded by offering a demonstration to explain how it all worked.

Sonnant styles itself as a “transformational artificial intelligence (AI) and machine learning (ML) company that provides content discovery for the spoken word”.

Simmons explained that the key to Sonnant’s success was their initial decision to train the AI to understand the Australian accent in phase one. “There’s often problems with the machine understanding the Aussie accent. Anyone who’s tried to make a booking at a restaurant while in America will know what I mean.”

Australian English is most associated with monophthongs (single vowels), where there are approximately 20 distinct sounds compared to American English, with only 16 sounds. Also difficult for the AI are Australian diphthongs, the timing between two vowel sounds and the tendency for a falling second sound.

An accurate transcript of an analogue recording is necessary to map the keywords. Simmons shows me how the platform extracts the keywords for use in various scenarios. Furthermore, users can set parameters or search for terms that the AI hasn’t selected.

Once a user has a summary of the terms, they can attach them to various curation problems. Let’s say SCA has noticed audiences downloading their morning shows late in the day, possibly listening while commuting home. The shows need to have commercials attached that are more appropriate for that time of day; delete the breakfast cereal and add commercials for Deliveroo.

By the end of the month, Sonnant will offer users the ability to curate their listening by using keywords and subjects.

Originally Sonnant was used in the education sector. Students and staff have been able to use the program to revise and redistribute lectures. The possibilities for students to curate course notes to reflect their research or to improve weaknesses in their knowledge is continually developing.

Simmons and I discuss the possibilities; I suggest having at your fingertips the ability to retrieve archival records of Dwight Eisenhower’s wartime comments on the war in Europe, against his domestic policy priorities while he was President in the ’50s.

The Future Is Here

Simmons leans over his desk and tells me he’s calling this ability to extract useable analogue recording as ‘archival revival’.

We return to the topic of SCA and how they might apply the platform to their industry. Simmons puts it in terms of “talent and topics”.

Radio audiences will either follow individual personalities or hone in on topics. Take sports radio. Audiences will want to curate their listening to suit their interests. An AFL demographic will have different consumer biases than a basketball listening demographic. There will be cross overs, but for SCA, the ability to tailor every aspect of the listening experience will allow them to extract maximum revenue.

Finally, if you can flexibly curate your audio via AI and machine learning, you probably need less expensive human staff in the production studio. Productivity will rise, leaving more income to drop to the bottom line.

As Simmons ends, he tells me that “the technology now allows us to do things with audio that was science fiction a few years ago. There’s plenty more to come.”

The post The Rise Of The Machines; Analogue Meets Artificial Intelligence appeared first on Artificial Intelligence.

AIDA Diagnostics presents an AI-powered smart assistant to enable crucial decision-making in blood transfusions.

Two young doctors set out with one dream: to help themselves and others to perform blood transfusions efficiently and safely. They began with a clinical decision support system and are now trying to digitalise the ‘from vein to vein’ concept, helping doctors, hospitals, and blood donation centres to develop a data-driven process for blood usage.

The AIDA Diagnostics project started in 2019, when co-founders Dr Mateusz Pawełczuk and Dr Michał Janiszewski observed problems with blood transfusion in their practice. Clinical guidance for transfusions, which included advice such as ‘If you want to raise the level of haemoglobin by one, use one full unit of blood’ and ‘If you want to use one unit of blood use two of them’, sounded more like proverb-based medicine than evidence-based practice. With some experience and general knowledge of Artificial Intelligence (AI) on their side, Pawełczuk and Janiszewski realised it was entirely feasible to create an AI-powered smart assistant capable of assisting in blood transfusions. This is how AIDA Diagnostics was born.

Clinical decision support system

After a long phase of data gathering, the AIDA Diagnostics team finally created a system which can assist with making the best possible decisions regarding blood transfusions. There were a few challenges along the way – the system does not provide full explanations for its decisions, meaning some of our experts did not wholly trust its findings – but with their help, we managed to create a hybrid system which benefits from neural networks within an expert system that controls the results of AI analysis. We can determine exactly how much blood should be used for each individual patient – of course we are currently limited by the volume of blood bags, but we expect this to be a temporary limitation. In addition to this information, we can tell our users what could be done with patients if they do not want to receive a blood transfusion: proposed alternatives could include further observation, fluid transfusion or potentially the administration of vasopressin analogues.

Because we care deeply about the welfare of patients, we want to continue to develop our medical device; and in order to do this we need to conduct further clinical trials of our innovative solution. The trials we had scheduled for 2020 were unfortunately postponed by the onset of the COVID-19 pandemic, but we tentatively expect trials to resume in the Autumn of 2021. If you are interested in participating in one of our trials, please get in touch by email.

Hospitals

Diving deep into the challenges facing blood management has given us a unique understanding of the ways in which the adoption and integration of technological solutions could help to address these challenges. The AIDA Blood: Hospitals system uses genetic algorithms to enable it to reach a high level of accuracy in its predictions. The system begins by considering the potential viability of several models; and determining the most appropriate next steps given the patient’s specific circumstances. By combining datasets on multiple treatment protocols, much like some cells combine DNA, the second generation is born. We repeat this process until we have the best model suited to each individual hospital.

AIDA’s data modelling constantly gathers new data from the hospital where it is installed, reaching its full potential after the first month of usage. Not only is the algorithm informed by historical data, but by integrating the hospital’s electronic health records (EHRs) it is able to adjust its projections in real time. As a result of this, the teams responsible for blood storage and management can make data-driven decisions with all necessary information already in one place. Our system is designed to give users as much control as possible, so each user can adjust parameters, such as the metrics for alerting them to the risk of insufficient quantities of blood, to their own needs.

Blood donation centres

Hospitals need to get their blood from blood donation facilities. These institutions are crucial elements of the whole blood industry, so we could not call AIDA Diagnostics a blood-focused startup if we did not take them into account. Our AIDA Blood: Blood Bank system works in a similar fashion to the Hospital system in terms of developing and presenting predictions. The only difference is in the types of real-time data we are using. One interesting factor on this front is that we can provide blood banks with recommendations on the best ways to communicate with their donors. With the help of our system, blood donation centres can move from the reactive method of inviting donors when stock levels are low towards the proactive approach, wherein stock levels remain at a broadly consistent level. In addition, the algorithms used by AIDA Blood: Blood Bank can help track and communicate with donors with rare blood types.

With the help of our algorithms, you can make your blood management smarter. You can accomplish this by installing our standalone software, or it can also be integrated into an existing solution.

The post AIDA Diagnostics: when blood meets Artificial Intelligence appeared first on Artificial Intelligence.