AAS 25-147 INVESTIGATING THE IMP ACT OF OBSERVATION SP ACE DESIGN CHOICES ON TRAINING REINFORCEMENT LEARNING SOLUTIONS FOR SP ACECRAFT PROBLEMS Nathaniel Hamilton *, Kyle Dunlap, and Kerianne L. Hobbs Recent research using Reinforcement Learning (RL) to learn autonomous control for spacecraft operations has shown great success. However, a recent study showed their performance could be improved by changing the action space, i.e. control outputs, used in the learning environment. This has opened the door for finding more improvements through further changes to the environment. The work in this paper focuses on how changes to the environment's observation space can impact the training and performance of RL agents learning the spacecraft inspection task. The studies are split into two groups. The first looks at the impact of sensors that were designed to help agents learn the task. The second looks at the impact of reference frames, reorienting the agent to see the world from a different perspective. The results show the sensors are not necessary, but most of them help agents learn more optimal behavior, and that the reference frame does not have a large impact, but is best kept consistent. INTRODUCTION Autonomous spacecraft operation is a critical capability for managing the growing number of space and increasingly complex operations.

Every few years, Hany Farid and his wife have the grim but necessary conversation about their end-of-life plans. They hope to have many more decades together--Farid is 58, and his wife is 38--but they want to make sure they have their affairs in order when the time comes. In addition to discussing burial requests and financial decisions, Farid has recently broached an eerier topic: If he dies first, would his wife want to digitally resurrect him as an AI clone? Farid, an AI expert at UC Berkeley, knows better than most that physical death and digital death are two different things. "My wife has my voice, my likeness, and a lot of my writings," he told me. "She could very easily train a large language model to be an interactive version of me."

A handful of important trends will transform the software developer experience in 2023, as enterprises consider more self-hosting, observe more SaaS consolidations and see an upswing of interest in creative AI. Also, as AI enters the creativity realm, it threatens to upend the future of app dev. And OpenAI's Chat GPT, released in November, takes code completion beyond line suggestions -- in addition to writing complete web pages and simple applications, it can generate new programming languages. For developers, the 2022 job market started strong, but by December, they saw storm clouds as layoffs hit the tech sector. Experts felt vibes of the early 2000s recession and the pandemic's early days.

Software trends come and go, but developers can future-proof their resumes with these major trends, industry experts say. Years ago, software developers could build a career on a single language, such as React.js, "Nowadays, there is so much blending between websites, e-commerce platforms, mobile applications, cloud and all the pieces in between that a developer has to learn multiple languages and frameworks," Schumann said. But while software development mainstays such as cloud computing and mobile apps are trends that endured, hundreds of others have faded into the annals of history, including LISP, marketing reporting software and storage tapes. This can make it a challenge for developers to decide which bandwagon to jump on.

Reinforcement Learning (RL) has become an increasingly important research area as the success of machine learning algorithms and methods grows. To combat the safety concerns surrounding the freedom given to RL agents while training, there has been an increase in work concerning Safe Reinforcement Learning (SRL). However, these new and safe methods have been held to less scrutiny than their unsafe counterparts. For instance, comparisons among safe methods often lack fair evaluation across similar initial condition bounds and hyperparameter settings, use poor evaluation metrics, and cherry-pick the best training runs rather than averaging over multiple random seeds. In this work, we conduct an ablation study using evaluation best practices to investigate the impact of run time assurance (RTA), which monitors the system state and intervenes to assure safety, on effective learning. By studying multiple RTA approaches in both on-policy and off-policy RL algorithms, we seek to understand which RTA methods are most effective, whether the agents become dependent on the RTA, and the importance of reward shaping versus safe exploration in RL agent training. Our conclusions shed light on the most promising directions of SRL, and our evaluation methodology lays the groundwork for creating better comparisons in future SRL work.

When Russia invaded Ukraine, the US Department of Defense turned to a team of machine learning and artificial intelligence experts to make sense of an avalanche of information about the conflict. "We have surged data scientists forward," Deputy Secretary of Defense Kathleen Hicks told WIRED in a recent interview. These tech experts crafted code and machine learning algorithms, creating systems that are "especially valuable for synthesizing the complex logistics picture," she said. Due to the sensitive nature of operations in Ukraine, Hicks says she cannot provide details of what the data team has done. But Hicks says this helps prove a point that she and others have been making within the Pentagon for some time--that technology is fundamentally changing the nature of war, and the US needs to adapt in order to maintain its edge.

Outgoing senior Pentagon official Preston Dunlap explains why he left the Pentagon over fears of U.S. losing its technological edge A top Pentagon official is stepping down, warning Tuesday the United States is falling behind and could lose its technological edge to adversaries like China. Preston Dunlap, the chief architect for the Space Force, argued the military used to excel in areas like artificial intelligence on "Fox & Friends First," but the commercial sector is now surpassing the defense community. He said this is ultimately providing an opportunity to U.S. adversaries. "These are accessible to anyone with resources and academics and capabilities, and so our adversaries or potential adversaries are able to have access to that technology, not only inside their own economies, but because of the benefit of our free and open society, which is a great thing," Dunlap told co-host Todd Piro. "They also have access to a lot of our capabilities as well from our companies, and so what I want us to be able to do is to make sure that we don't just compete globally on that technological scale, but we can actually adapt and adopt the technologies of our own companies and commercial ecosystem, which we have the opportunity to do right here at home," he continued.

A senior official responsible for driving technological innovation at the U.S. Department of Defense has resigned, saying the Pentagon needs "structural change" and should behave more like SpaceX, Elon Musk's satellite company that has shaken up rocket launches. "We're falling behind the commercial base in key areas, so we've got to catch up," Preston Dunlap, the first person in the U.S. Department of Defense to fulfill the role of chief architect officer, said in an interview. As a result the U.S. risked losing its technological edge against potential adversaries, he said. Dunlap, who handed in his resignation on Monday after three years in the post at the U.S. Space Force and U.S. Air Force, was responsible for pushing more technology into a $70 billion budget for research, development and acquisition. He plans to start a space software company focused on the nexus with satellites, data and artificial intelligence.

Investment in artificial intelligence startups hit a record high of $59 billion in 2021, up from $28 billion in 2020, according to Crunchbase data. Venture investment in AI startups will only continue to rise in the next five years. That's because AI has finally reached a tipping point where it is powerful and affordable enough to make a real economic impact on diverse industries. AI is a broad category, but generally encompasses all "intelligent" software that can learn from itself in some capacity. This can include machine-learning platforms and the software used by robots, and self-driving vehicles to interpret the environments around them.

Conversations with chatbots are getting a lot more interesting. Fueled by artificial intelligence and natural language processing, chatbot technology has taken leaps forward over the past few years. "Chatbots are now a critical component in the customer omnichannel experience," said Rob Dunlap, Associate Partner – Cognitive & Analytics Practice Leader with IBM at a recent CanadianCIO Virtual Roundtable. In the past, the introduction of chatbots hasn't always delivered better customer service. However, the technology is changing rapidly to meet consumer needs, said Dunlap.