Training generalist robot agents is an immensely difficult feat due to the requirement to perform a huge range of tasks in many different environments. We propose selectively training robots based on end-user preferences instead. Given a factory model that lets an end user instruct a robot to perform lower-level actions (e.g. 'Move left'), we show that end users can collect demonstrations using language to train their home model for higher-level tasks specific to their needs (e.g. 'Open the top drawer and put the block inside'). We demonstrate this hierarchical robot learning framework on robot manipulation tasks using RLBench environments. Our method results in a 16% improvement in skill success rates compared to a baseline method. In further experiments, we explore the use of the large vision-language model (VLM), Bard, to automatically break down tasks into sequences of lower-level instructions, aiming to bypass end-user involvement. The VLM is unable to break tasks down to our lowest level, but does achieve good results breaking high-level tasks into mid-level skills. We have a supplemental video and additional results at talk-through-it.github.io.

Machine learning falls into a category of technology currently experiencing hyper-exponential growth as enterprises capitalize on its ability to transform data into insightful action. Like any hyped technology, machine learning is not without current limitations; however, companies operating on the cutting-edge are finding innovative ways to integrate machine learning into an impressive bottom-line. And it's no longer just pet projects for elite Fortune 500 brands--everyone is joining the fun. With this rapid evolution, the role of data professionals is being reconceptualized; leaders increasingly understand data-related infrastructure in terms of the factory model, giving rise to the notion of a data science factory. Data goes in, actionable insights come out, and everything happening in-between falls into the "making the sausage" category--nobody wants to know too much. On one end of the factory, you have the dizzying collection of platforms to manipulate data (Python, H2O, TensorFlow, R, Scikit-Learn, Keras, SAS, Openface, Caffe2, Watson, Google, Azure, AWS ML cloud APIs, and the list goes on).

There has been a lot of excited talk recently about the threat to jobs posed by automation, robots, and now Artificial Intelligence (AI): machines that can think like humans. We're told that ever more complex tasks can now be automated and perhaps done better as a result, and we should all be preparing for a world in which we're competing for work with computers. Is teaching one of the jobs put at risk by the emergence of AI? Or does AI have potential to enhance life in the classroom? A recent event organised by BESA, the industry body for education suppliers, provided plenty of food for thought about these questions. He argued that AI can help us move away from the "factory model of education" towards a more open-ended system focused on creativity and problem solving – and he said we're seeing early signs of what technology can bring us in innovations such as "no lecture hall" universities and courses offering "nanodegrees".

Advanced analytics techniques, such as artificial intelligence and machine learning, provide organizations with new insights not possible with traditional analytics. To take advantage of these technologies and drive competitive advantage, organizations need to design and build solutions that allow them to exponentially grow their capacity to create value from data. The challenge is, how do you do that without also exponentially growing infrastructure costs and the number of data scientists needed to meet that business demand? The answer lies in industrializing the process using a data factory model. As AI / ML technologies, packaging, frameworks and tooling are emerging rapidly, there is a real need to evaluate these new capabilities to understand the potential impact they might have on your business.

We've understood for a long time that organizations who spend more on, and are better at, deriving value from their data using analytics significantly outperform their peers in the market. All of us also know, because we feel it, that the pace of change is ever increasing. I see this all the time with the customers I work with, many of whom seem to be suffering from the "Red Queen" effect – each having to change and innovate faster just to keep standing still, let alone make progress against a tide of change. I've also had cause to re-read Salim Ismail's book, "Exponential Organizations", recently which got me thinking: in order to compete, we should be designing and building solutions that allow us to exponentially grow our capacity to create value from data in order to meet business demand. Importantly though, how do we do that without also exponentially growing infrastructure costs (a bad idea) and the number of Data Scientists employed (an impossible dream)?

We've understood for a long time that organizations who spend more on, and are better at, deriving value from their data using analytics significantly outperform their peers in the market. All of us also know, because we feel it, that the pace of change is ever increasing. I see this all the time with the customers I work with, many of whom seem to be suffering from the "Red Queen" effect – each having to change and innovate faster just to keep standing still, let alone make progress against a tide of change. I've also had cause to re-read Salim Ismail's book, "Exponential Organizations", recently which got me thinking: in order to compete, we should be designing and building solutions that allow us to exponentially grow our capacity to create value from data in order to meet business demand. Importantly though, how do we do that without also exponentially growing infrastructure costs (a bad idea) and the number of Data Scientists employed (an impossible dream)?