We evaluate davinci-003 across a range of classification, summarization, and generation tasks. Using Scale Spellbook, the platform for large language model apps, we show where davinci-003 significantly outperforms the prior version and where it still has room to improve.
Character is a full stack Artificial General Intelligence (AGI) company.
What if you could create your own AI, and it was always available to help you with anything?
Imagine everything it could do for you, from being your own personal teacher, assistant or even friend. Two months after launching in September, our beta generates 1 billion words per day. Get a glimpse of the future at character.ai.
Researchers at Worcester Polytechnic Institute recently created OmniWheg, a robotic system that can adapt its configuration while navigating its surrounding environment, seamlessly changing from a wheeled to a legged robot. This robot, introduced in an IEEE IROS 2022 paper, pre-published on arXiv, is based on an updated version of the so-called “whegs,” a series of mechanisms design to transform a robot’s wheels or wings into legs.
“Quadruped and biped robots have been growing in popularity, and the reason for that might be the search for ‘anthropomorphization’ that the general audience commonly engages in,” Prof. Andre Rosendo, one of the researchers who developed the robot, told TechXplore. “While ‘being capable of going everywhere we go’ sounds like an exciting appeal, the energetic cost of legs is very high. We humans have legs because that is what evolution gave us, but we wouldn’t dare to create a ‘legged car,’ as we know that this ride wouldn’t be as comfortable or energy efficient as a wheeled car ride.”
Continue reading “A transformable robot with an omnidirectional wheel-leg” »
Researchers at Tohoku University, the University of Messina, and the University of California, Santa Barbara (UCSB) have developed a scaled-up version of a probabilistic computer (p-computer) with stochastic spintronic devices that is suitable for hard computational problems like combinatorial optimization and machine learning.
Moore’s law predicts that computers get faster every two years because of the evolution of semiconductor chips. While this is what has historically happened, the continued evolution is starting to lag. The revolutions in machine learning and artificial intelligence means much higher computational ability is required. Quantum computing is one way of meeting these challenges, but significant hurdles to the practical realization of scalable quantum computers remain.
A p-computer harnesses naturally stochastic building blocks called probabilistic bits (p-bits). Unlike bits in traditional computers, p-bits oscillate between states. A p-computer can operate at room-temperature and acts as a domain-specific computer for a wide variety of applications in machine learning and artificial intelligence. Just like quantum computers try to solve inherently quantum problems in quantum chemistry, p-computers attempt to tackle probabilistic algorithms, widely used for complicated computational problems in combinatorial optimization and sampling.
Researchers have used a quantum processor to make microwave photons uncharacteristically sticky. They coaxed them to clump together into bound states, then found that these photon clusters survived in a regime where they were expected to dissolve into their usual, solitary states. The discovery was first made on a quantum processor, marking the growing role that these platforms are playing in studying quantum dynamics.
Photons—quantum packets of electromagnetic radiation like light or microwaves—typically don’t interact with one another. Two crossed flashlight beams, for example, pass through one another undisturbed. But in an array of superconducting qubits, microwave photons can be made to interact.
In “Formation of robust bound states of interacting photons,” published today in Nature, researchers at Google Quantum AI describe how they engineered this unusual situation. They studied a ring of 24 superconducting qubits that could host microwave photons. By applying quantum gates to pairs of neighboring qubits, photons could travel around by hopping between neighboring sites and interacting with nearby photons.
Good Morning, 2033 — A Sci-Fi Short Film.
What will your average morning look like in 2033? And who hacked us?
A grand vision in robot learning, going back to the SHRDLU experiments in the late 1960s, is that of helpful robots that inhabit human spaces and follow a wide variety of natural language commands. Over the last few years, there have been significant advances in the application of machine learning (ML) for instruction following, both in simulation and in real world systems. Recent Palm-SayCan work has produced robots that leverage language models to plan long-horizon behaviors and reason about abstract goals. Code as Policies has shown that code-generating language models combined with pre-trained perception systems can produce language conditioned policies for zero shot robot manipulation. Despite this progress, an important missing property of current “language in, actions out” robot learning systems is real time interaction with humans.
Ideally, robots of the future would react in real time to any relevant task a user could describe in natural language. Particularly in open human environments, it may be important for end users to customize robot behavior as it is happening, offering quick corrections (“stop, move your arm up a bit”) or specifying constraints (“nudge that slowly to the right”). Furthermore, real-time language could make it easier for people and robots to collaborate on complex, long-horizon tasks, with people iteratively and interactively guiding robot manipulation with occasional language feedback.
Women are playing a significant role in the field of Artificial Intelligence. Take a look at the well-known women personalities in the AI field here.
How far away could an artificial brain be? Perhaps a very long way off still, but a working analogue to the essential element of the brain’s networks, the synapse, appears closer at hand now.
That’s because a device that draws inspiration from batteries now appears surprisingly well suited to run artificial neural networks. Called electrochemical RAM (ECRAM), it is giving traditional transistor-based AI an unexpected run for its money—and is quickly moving toward the head of the pack in the race to develop the perfect artificial synapse. Researchers recently reported a string of advances at this week’s IEEE International Electron Device Meeting (IEDM 2022) and elsewhere, including ECRAM devices that use less energy, hold memory longer, and take up less space.
The artificial neural networks that power today’s machine-learning algorithms are software that models a large collection of electronics-based “neurons,” along with their many connections, or synapses. Instead of representing neural networks in software, researchers think that faster, more energy-efficient AI would result from representing the components, especially the synapses, with real devices. This concept, called analog AI, requires a memory cell that combines a whole slew of difficult-to-obtain properties: it needs to hold a large enough range of analog values, switch between different values reliably and quickly, hold its value for a long time, and be amenable to manufacturing at scale.
