Tag: ai

DARPA $2B AI Campaign

To address the limitations of the first and second wave AI technologies, DARPA seeks to explore new theories and applications that could make it possible for machines to adapt to changing situations. DARPA sees this next generation of AI as a third wave of technological advance, one of contextual adaptation. To better define a path forward, DARPA is announcing today a multi-year investment of more than $2B in new and existing programs called the “AI Next” campaign.

Hypocralypse?

Better tech for reading feelings and widespread hypocrisy, seem to me to be on a collision course. As a result, within a few decades, we may see something of a “hypocrisy apocalypse”, or “hypocralypse”, wherein familiar ways to manage hypocrisy become no longer feasible, and collide with common norms, rules, and laws. In this post I want to outline some of the problems we face.

General Evolvable Brains

Those who are trying to improve such systems have long wondered: what is the secret of human general intelligence? In this post I want to consider we can learn about this from fact that the brain evolved. How would an evolved brain be general? if we are looking to explain a surprising generality, flexibility, and rapid evolution in human brains, it makes sense to consider the possibility that human brain design took a different path, one more like that of single-celled metabolism. That is, 1 straightforward way to design a general evolvable brain is to use an extra large toolbox of mental modules that can be connected together in many different ways. While each tool might be a carefully constructed jewel, the whole set of tools would have less of an overall structure. Like a pile of logical gates that can be connected many ways, or metabolism sub-networks that can be connected together into many networks. In this case, the secret to general evolvable intelligence would be less in the particular tools and more in having an extra large set of tools, plus some simple general ways to search in the space of tool combinations. A tool set so large that the brain can do most tasks in a great many different ways.

2023-03-25: Intelligence is modular and extremely prevalent, for generous definitions of intelligence

One implication of this hierarchy of homeostatically stable, nested modules is that organisms became much more flexible while still maintaining a coherent ‘self’ in a hostile world. Evolution didn’t have to tweak everything at once in response to a new threat, because biological subunits were primed to find novel ways of compensating for changes and functioning within altered systems. For example, in planarian flatworms, which reliably regenerate every part of the body, using drugs to shift the bioelectrically stored pattern memory results in two-headed worms. Remarkably, fragments of these worms continue to regenerate two heads in perpetuity, without editing the genome. Moreover, flatworms can be induced, by brief modulation of the bioelectric circuit, to regrow heads with shape (and brain structure) appropriate to other known species of flatworms (at about 100 million years of evolutionary distance), despite their wild-type genome.

AI Sector Blurring

As the applications of machine learning grow, the interactions between companies and nation states will grow in complexity. Consider for example road transportation, where we are gradually moving towards on demand, autonomous cars. This will increasingly blur the line between publicly funded mass transportation and private transport. If this leads to a new natural monopoly in road transportation should it be managed by the state or by a British company, or by a multinational company like Uber?

Low-light image enhancement

Imaging in low light is challenging due to low photon count and low SNR. Short-exposure images suffer from noise, while long exposure can induce blur and is often impractical. A variety of denoising, deblurring, and enhancement techniques have been proposed, but their effectiveness is limited in extreme conditions, such as video-rate imaging at night. To support the development of learning-based pipelines for low-light image processing, we introduce a dataset of raw short-exposure low-light images, with corresponding long-exposure reference images. Using the presented dataset, we develop a pipeline for processing low-light images, based on end-to-end training of a fully-convolutional network. The network operates directly on raw sensor data and replaces much of the traditional image processing pipeline, which tends to perform poorly on such data. We report promising results on the new dataset, analyze factors that affect performance, and highlight opportunities for future work.

In silico labeling

The new deep-learning network can identify whether a cell is alive or dead, and get the answer right 98% of the time (humans can typically only identify a dead cell with 80% accuracy) — without requiring invasive fluorescent chemicals, which make it difficult to track tissues over time. The deep-learning network can also predict detailed features such as nuclei and cell type (such as neural or breast cancer tissue).

China AI Ethical Issues

Many, though not all, of these new surveillance technologies are powered by AI. Recent advances in AI have given computers superhuman pattern-recognition skills: the ability to spot correlations within oceans of digital data, and make predictions based on those correlations. It’s a highly versatile skill that can be put to use diagnosing diseases, driving cars, predicting consumer behavior, or recognizing the face of a dissident captured by a city’s omnipresent surveillance cameras. The Chinese government is going for all of the above, making AI core to its mission of upgrading the economy, broadening access to public goods, and maintaining political control.