Samsung launches Bixby (and the Galaxy S8), DeepMind tackles an AI pain point and Elon Musk unveils his new venture
Posted by Jeevan Jayaprakash
In this issue, we take a look at the launch of Bixby, DeepMind’s recent accomplishment (yes, they never let up) and the next big thing in Elon Musk’s crosshairs.
If you have any general questions about the contents of our digests or what it could mean for you, please don’t hesitate to drop us a message.
Have an awesome weekend,
Hi Mum! Said Dad.
Samsung launches Bixby (and the Galaxy S8)
The Galaxy S8 was unveiled on Wednesday at simultaneous events held in New York and London. Admittedly, unveil is probably the wrong word to use as there wasn’t much left to unveil following the barrage of leaks in the run up to the showpiece event.
Whilst Samsung has again laid down the gauntlet in terms of design with its bezel-less HDR display, dubbed the ‘infinity display’, it was the inclusion of Bixby, Samsung’s rival to Alexa/Siri/Google’s Assistant which caught our eye. This was an admission by another tech behemoth that voice is one of the leading frontrunners for the next big platform shift (which is of course inextricably linked to the advancements in AI — the real MVP) because let’s face it, smartphones are no longer a growth area — they are ubiquitous. Instead, this ubiquity means they are increasingly beginning to look like a vessel for something else. Whether that is voice, augmented reality or something else altogether remains to be seen.
Considering that the Google Assistant will be baked into the Galaxy S8 by default as part of Android, it is telling that Samsung felt the need to create their own assistant. Samsung clearly wants to own and dictate the experience for customers and not leave this to Google. What we must also remember is that Samsung, unlike Google/Amazon and Apple to a lesser extent, is a hardware manufacturer through and through. Samsung manufactures TVs, washing machines and fridges, all of which Samsung has said will become internet connected by 2020. You can now begin to see the rationale for Bixby.
Whilst Samsung may not have the wealth of data that a Google or an Amazon has accumulated, it has reach by virtue of being one of the world’s leading appliance manufacturers. If their claims of Bixby becoming a leader in natural language processing (with the help of Viv Labs) comes to fruition, Bixby will be a formidable adversary in the race to own the connected home.
DeepMind claims to have given algorithms the ability to remember
Catastrophic forgetting.
This is the term used by scientists to describe the phenomenon whereby neural networks (the highly in vogue algorithms, loosely based on the human brain, that are used by the likes of Amazon’s Alexa to work out how to respond to what we say and Facebook to automatically recognise and suggest friends to tag in your photos) ‘forgets’ what it has learned previously when it is taught something new. You might think that sounds a bit contradictory — don’t all these companies say that their assistants/services become ‘smarter’ the more you use them? That is certainly true but only within the context of one task. However, DeepMind, the artificial intelligence company owned by Google’s parent company, Alphabet Inc., claims to have overcome this sticking point. The implication of such a breakthrough is that it means an AI system can be applied to a range of tasks rather than just being proficient in one narrowly defined task.
The researchers at DeepMind have coined their algorithm ‘Elastic Weight Consolidation’ and the way it works is as follows: when the algorithm is trained to undertake task A, it assigns a weighting to the nodes which are considered to be important for undertaking task A (almost like how our brain becomes familiar with a new skill), and when subsequently trained on task B, the weightings assigned to these ‘important’ nodes slows down the rate at which the value of these nodes can be altered — in other words, they don’t forget whilst being trained to undertake a new task.
Why is this important? Well, the pinnacle of the field of AI is considered to be the achievement of artificial general intelligence (AGI) — a general purpose system with intelligence that is comparable to the human mind and possibly even well beyond human capabilities. In order to achieve AGI, the ability to learn sequentially goes without saying so DeepMind’s work is a step in the right direction. However, it must be added that whilst this work is certainly exciting, we are still some way away from achieving AGI and all the commercial applications that will follow.
Elon Musk sets out his ambition to connect our brains to computers
He’s built PayPal, he’s built a space company (to say he has taught himself rocket science is no exaggeration considering he serves as CTO) and an electric car company quite literally from the ground up and he recently founded a non-profit research company to help democratise AI.
A maverick amongst mavericks. Musk has always had a penchant for going after the toughest of problems in order to advance human welfare and he’s just founded another company to do just that: Neuralink.
Neuralink’s aim is to develop ‘neural lace’ technology which will help to implant tiny electrodes into the brain. These electrodes could be used to improve memory or augment human intelligence with artificial intelligence (also referred to as ‘brain hacking’). This is something that Musk has openly talked about for a while now and he has even explicitly mentioned the idea of “uploading our brains to the cloud”. Some have speculated that this is Musk’s attempt to ensure humanity can keep up with the progress of AI. Sound crazy? Well, let’s not forget that people called him crazy when he said he was going to build rockets to help us to colonise Mars one day (admittedly, he hasn’t achieved this yet but he’s got a lot further than anyone expected and more importantly, he’s still in the game!) and when he said he was going to build autonomous electric cars from scratch.
According to the Wall Street Journal, leading academics in the field have bought into Musk’s vision and signed up to join Neuralink. The hope is that with the Musk’s financial backing, academics may be able to expedite the process for making this a reality. In typically daredevil fashion, Musk has set out a timeline of 4 to 5 years for achieving improvements in cognitive function but Blake Richards, a neuroscientist and assistant professor at the University of Toronto has said that this is a “few decades down the road” but that it is “certainly possible” within the 21st century.
Well, all the best Elon — you are going to need it if you are to enhance human cognition alongside building safe, affordable autonomous electric cars and reusable, self-landing rockets.
Originally written as part of Hi Mum! Said Dad’s Weekly Digest.
If you would like to receive our Weekly Digest straight to your inbox, please drop us an email at hello@himumsaiddad.com.
