This is the third SyncDevelopHER event and Vickie lined up two great women speakers. It started off with Lily Ash Sakula who is a Partner at Bethnal Green Ventures telling us how they fund startups for three months with £15,000 in turn for a 6% share. She suggested that the technology world is probably less equal than others in many ways, that it’s bad that conference panels often only feature men and that only a tiny percentage of female founded startups get funding. While this is all true, in my opinion, it’s not for want of encouragement.
Lily then went through the usual ‘lady geek’ type arguments about why women don’t feel comfortable in some tech environments and why many are not interested in getting into tech. As I am a man you may feel that my opinion on this has little worth, but I don’t agree with a lot of these arguments and I think we need a fresh perspective and a practical, deliverable plan of action if we want to encourage more women into tech. Although I feel the the reasons why we want more women in tech are becoming clearer, at least to me, and have moved beyond the ‘lady geek’ attitude that “it’s just not fair that there aren’t”, the how still needs more attention.
Lily went on to explain that if men prevail as the main problem solvers they will continue solving male problems. Diversity is important. She showed us a hilarious app that demonstrated the issue well. It was ever so slightly risque, so I won’t mention here, but you’re welcome to ask me about it. It was pointed out from the audience that women participate in similar behaviour to that promoted by the app, which readdressed the balance slightly, but the point still very much stands.
To finish up, Lily told us about a group called TechMums run by one of the saviours of Bletchley Park, Sue Black. It is intended to build confidence in programming and to encourage it's members to help their kids.
Catherine Breslin gave most of her session with her baby strapped to her chest, which was absolutely brilliant! Catherine has worked in speech technology for ten years and has carried out a lot of research to improve speech technology. She has a degree from Oxford, a PHD from Cambridge and worked for Toshiba before taking a career break to have children. Next year she’ll start working for Amazon on a top secret project that not even she knows about yet (voice operated drones anyone?).
Catherine started off by taking us through the history of machine learning including the Turing test. She showed us ‘real big data’ on a graph that showed the amount of data in the world now and in the future in zettabytes (one billion terabytes). The contributing sources include smart phones, computers and social media. Most of this data is transient and not stored. In the future when storage is even cheaper, less will be thrown away - a scary thought!
Catherine went on to explain that this amount of data cannot be looked at manually. Traditionally you decide what your computer program is going to do and then write and execute it, but in this instance the amounts of data are too big and too varied for this to work. Machine learning is ideal for analysing large amounts of varied data, but because it’s grounded in probability it makes testing and debugging quite difficult.
Catherine gave us spam filtering as an example of machine learning. Emails are used to train a spam filter. When a new email arrives the system must decide if is it spam or not. Humans marking emails as spam helps the machine learn. There are a number of libraries available, including one written in Python, for machine learning.
Object recognition, sentiment analysis (positive or negative) and fraudulent transactions are areas where machine learning is applied. Catherine is particularly interested in audio. Machine learning can be used to help identify gender, identify adverts, separate speech and for translation.
Spoken languages can be very difficult to model. However, machine learning can be used to identify likely word sequences in a model and then used to determine what is said in audio. Other factors such as background noise can make it difficult for a machine to understand speech. Applications include assisting technology, helping people to speak. helping the elderly and hard of hearing, in car control and preserving languages.
Although there is a date (13th March), there is no agenda for the next SyncDevelopHER, but I can’t wait! If the speakers are anything like tonight’s pair it will be fantastic.
Also published on the Norfolk Tech Journal.
Comments
Post a Comment