Learning few-shot imitation as cultural transmission.

Cultural transmission is the domain-general social skill that allows agents to acquire and use information from each other in real-time with high fidelity and recall. It can be thought of as the process that perpetuates fit variants in cultural evolution. In humans, cultural evolution has led to the accumulation and refinement of skills, tools and knowledge across generations. We provide a method for generating cultural transmission in artificially intelligent agents, in the form of few-shot imitation. Our agents succeed at real-time imitation of a human in novel contexts without using any pre-collected human data. We identify a surprisingly simple set of ingredients sufficient for generating cultural transmission and develop an evaluation methodology for rigorously assessing it. This paves the way for cultural evolution to play an algorithmic role in the development of artificial general intelligence.

Communicative capital: a key resource for human-machine shared agency and collaborative capacity.

In this work, we present a perspective on the role machine intelligence can play in supporting human abilities. In particular, we consider research in rehabilitation technologies such as prosthetic devices, as this domain requires tight coupling between human and machine. Taking an agent-based view of such devices, we propose that human-machine collaborations have a capacity to perform tasks which is a result of the combined agency of the human and the machine. We introduce communicative capital as a resource developed by a human and a machine working together in ongoing interactions. Development of this resource enables the partnership to eventually perform tasks at a capacity greater than either individual could achieve alone. We then examine the benefits and challenges of increasing the agency of prostheses by surveying literature which demonstrates that building communicative resources enables more complex, task-directed interactions. The viewpoint developed in this article extends current thinking on how best to support the functional use of increasingly complex prostheses, and establishes insight toward creating more fruitful interactions between humans and supportive, assistive, and augmentative technologies.

Reactive Reinforcement Learning in Asynchronous Environments.

The relationship between a reinforcement learning (RL) agent and an asynchronous environment is often ignored. Frequently used models of the interaction between an agent and its environment, such as Markov Decision Processes (MDP) or Semi-Markov Decision Processes (SMDP), do not capture the fact that, in an asynchronous environment, the state of the environment may change during computation performed by the agent. In an asynchronous environment, minimizing reaction time-the time it takes for an agent to react to an observation-also minimizes the time in which the state of the environment may change following observation. In many environments, the reaction time of an agent directly impacts task performance by permitting the environment to transition into either an undesirable terminal state or a state where performing the chosen action is inappropriate. We propose a class of reactive reinforcement learning algorithms that address this problem of asynchronous environments by immediately acting after observing new state information. We compare a reactive SARSA learning algorithm with the conventional SARSA learning algorithm on two asynchronous robotic tasks (emergency stopping and impact prevention), and show that the reactive RL algorithm reduces the reaction time of the agent by approximately the duration of the algorithm's learning update. This new class of reactive algorithms may facilitate safer control and faster decision making without any change to standard learning guarantees.

Differential Responses of Post-Exercise Recovery of Leg Blood Flow and Oxygen Uptake Kinetics in HFpEF versus HFrEF.

The goals of the current study were to compare leg blood flow, oxygen extraction and oxygen uptake (VO2) after constant load sub-maximal unilateral knee extension (ULKE) exercise in patients with heart failure with reduced ejection fraction (HFrEF) compared to those with preserved ejection fraction (HFpEF). Previously, it has been shown that prolonged whole body VO2 recovery kinetics are directly related to disease severity and all-cause mortality in HFrEF patients. To date, no study has simultaneously measured muscle-specific blood flow and oxygen extraction post exercise recovery kinetics in HFrEF or HFpEF patients; therefore it is unknown if muscle VO2 recovery kinetics, and more specifically, the recovery kinetics of blood flow and oxygen extraction at the level of the muscle, differ between HF phenotypes. Ten older (68±10yrs) HFrEF (n = 5) and HFpEF (n = 5) patients performed sub-maximal (85% of maximal weight lifted during an incremental test) ULKE exercise for 4 minutes. Femoral venous blood flow and venous O2 saturation were measured continuously from the onset of end-exercise, using a novel MRI method, to determine off-kinetics (mean response times, MRT) for leg VO2 and its determinants. HFpEF and HFrEF patients had similar end-exercise leg blood flow (1.1±0.6 vs. 1.2±0.6 L/min, p>0.05), venous saturation (42±12 vs. 41±11%, p>0.05) and VO2 (0.13±0.08 vs. 0.11±0.05 L/min, p>0.05); however HFrEF had significantly delayed recovery MRT for flow (292±135sec. vs 105±63sec., p = 0.004) and VO2 (95±37sec. vs. 47±15sec., p = 0.005) compared to HFpEF. Impaired muscle VO2 recovery kinetics following ULKE exercise differentiated HFrEF from HFpEF patients and suggests distinct underlying pathology and potential therapeutic approaches in these populations.

Feasibility and reproducibility of measurement of whole muscle blood flow, oxygen extraction, and VO2 with dynamic exercise using MRI.

PURPOSE: Develop an MRI method to estimate skeletal muscle oxygen consumption (VO2 ) with dynamic exercise using simultaneous measurement of venous blood flow (VBF) and venous oxygen saturation (SvO2 ). METHODS: Real-time imaging of femoral VBF using a complex-difference method was interleaved with imaging of venous hemoglobin oxygen saturation (SvO2 ) using magnetic susceptometry to estimate muscle VO2 (Fick principle). Nine healthy subjects performed repeated 5-watt knee-extension (quadriceps) exercise within the bore of a 1.5 Tesla MRI scanner, for test/re-test comparison. VBF, SvO2 , and derived VO2 were estimated at baseline and immediately (<1 s) postexercise and every 2.4 s for 4 min. RESULTS: Quadriceps muscle mass was 2.43 ± 0.31 kg. Mean baseline values were VBF = 0.13 ± 0.06 L/min/kg, SvO2 = 69.4 ± 10.1%, and VO2 = 6.8 ± 4.1 mL/min/kg. VBF, SvO2 , and VO2 values from peak exercise had good agreement between trials (VBF = 0.9 ± 0.1 versus 1.0 ± 0.1 L/min/kg, R(2) = 0.83, CV = 7.6%; SvO2 = 43.2 ± 13.5 versus 40.9 ± 13.1%, R(2) = 0.88, CV = 15.6%; VO2 = 95.7 ± 18.0 versus 108.9 ± 17.3 mL/min/kg, R(2) = 0.88, CV = 12.3%), as did the VO2 recovery time constant (26.1 ± 3.5 versus 26.0 ± 4.0 s, R(2) = 0.85, CV = 6.0%). CV = coefficient of variation. CONCLUSION: Rapid imaging of VBF and SvO2 for the estimation of whole muscle VO2 is compatible with dynamic exercise for the estimation of peak values and recovery dynamics following exercise with good reproducibility.