Lifeboat Foundation

The latest language models are amazing. But for general artificial intelligence, I suspect we’re barking up the wrong tree.

It’s odd to try to create artificial intelligence without taking inspiration from the only existing natural intelligences we have: humans and animals. Yet that is the current direction of AI research.

Engineers have created intelligent 3D printers that can quickly detect and correct errors, even in previously unseen designs, or unfamiliar materials like ketchup and mayonnaise, by learning from the experiences of other machines.

The engineers, from the University of Cambridge, developed a machine learning algorithm that can detect and correct a wide variety of different errors in real time, and can be easily added to new or existing machines to enhance their capabilities. 3D printers using the algorithm could also learn how to print new materials by themselves. Details of their low-cost approach are reported in the journal Nature Communications.

3D printing has the potential to revolutionize the production of complex and customized parts, such as aircraft components, personalized medical implants, or even intricate sweets, and could also transform manufacturing supply chains. However, it is also vulnerable to production errors, from small-scale inaccuracies and mechanical weaknesses through to total build failures.

It’s been a while since Elon Musk published an extensive blog post outlining his stance on a specific topic. On the official Tesla website, his last blog post was on August 24, 2018, when he explained his decision to keep Tesla a publicly-traded company. Fortunately, a new Elon Musk essay has been posted in China, outlining the Tesla CEO’s thoughts on a number of topics — from sustainability, the Tesla Bot’s real-world use, Neuralink’s focus on the disabled, and SpaceX’s exploration aspirations.

The new Elon Musk essay was published in China Cyberspace 0, the Cyberspace Administration of China’s (CAC) flagship magazine. A translation of the essay was posted by Yang Liu, a journalist from the state-owned news agency Xinhua 0, on the Beijing Channel blog. As could be seen in Liu’s post, Musk actually discussed a number of topics in detail.

In a way, the publication of the new Elon Musk essay in the CAC’s flagship magazine is significant. As noted by The Register 0, Musk’s essay suggests that Chinese authorities approve of the Tesla CEO’s positions on the topics he discussed. Only a few other foreign entrepreneurs would likely be given the same honor.

Simply put, we need a reliable and secure energy grid. Two ways to ensure continuous electricity regardless of the weather or an unforeseen event are by using distributed energy resources (DER) and microgrids. DER produce and supply electricity on a small scale and are spread out over a wide area. Rooftop solar panels, backup batteries, and emergency diesel generators are examples of DER. While traditional generators are connected to the high-voltage transmission grid, DER are connected to the lower-voltage distribution grid, like residences and businesses are.

Microgrids are localized electric grids that can disconnect from the main grid to operate autonomously. Because they can operate while the main grid is down, microgrids can strengthen grid resilience, help mitigate grid disturbances, and function as a grid resource for faster system response and recovery.

On June 27th 2022, Yann LeCun, one of the godfathers of artificial intelligence and Head of AI at Meta released his vision on how to build autonomous AI systems. Here is the link to the paper.

First of all, I really suggest you to read this paper. As mentioned in the prologue, the text is written with as little jargon as possible. It uses as little mathematical prior knowledge as possible to appeal to readers with various backgrounds. It’s essentially a vision of what might direct the research efforts at Meta and elsewhere in the industry.

When you start reading the paper, quite quickly, you realize that this vision is very ambitious and futuristic. After all, Yann is describing an autonomous and polyvalent AI system.

An artificial intelligence (AI) algorithm to detect subtle brain abnormalities that cause epileptic seizures has been developed. The abnormalities, known as focal cortical dysplasias (FCDs), can often be treated with surgery but are difficult to visualize on an MRI. The new algorithm is expected to give physicians greater confidence in identifying FCDs in patients with epilepsy.

The work, which was part of the Multicentre Epilepsy Lesion Detection (MELD) project, appeared in Brain “Interpretable surface-based detection of focal cortical dysplasias: a Multi-centre Epilepsy Lesion Detection study.” Konrad Wagstyl, PhD, and Sophie Adler, PhD, both from University College London, led an international team of researchers on the work.

To develop the algorithm, the team quantified features of the brain cortex—such as thickness and folding—in more than 1,000 patient MRI scans from 22 epilepsy centers around the world. They then trained the algorithm on examples labeled by expert radiologists as either being healthy or having FCD.

To evaluate the accuracy of the models²⁸, we sampled from the BNN or deep ensemble to determine their uncertainty predictions (Extended Data Fig. 3a, b). Correct predictions are oriented toward the lower and higher range of the output, representing greater certainty about samples’ states, whereas incorrect predictions tend towards the 0.5 threshold. We can therefore assume higher confidence in a model’s predictions by removing the predictions in the middle using thresholds. We evaluated a range of thresholds with several models (Extended Data Fig. 3c–f), which show a substantial increase in accuracy due to the ambiguous samples being discarded, including the ensemble of normalized models reaching accuracy of 97.2%. A similar approach was applied to other models, including the IR and RS models (Extended Data Fig. 3g, h), raising accuracy by 10–15%, although this reduces the number of cells considered.

To better understand the development of the senescent phenotype and how nuclear morphology changes over time, we analyzed human fibroblasts induced to senescence by 10 Gy IR and imaged at days 10, 17, 24 and 31. The predictor identifies senescence at all four times points with probability that increases from days 10 to 17 but declines by day 31 (Extended Data Fig. 4a). Interestingly, examining the probability distribution of the predictor it was apparent that a growing peak of nonsenescent cells appear after day 17, suggesting that a small number of cells were able to escape senescence induction and eventually overgrow the senescent cells (Extended Data Fig. 4b). Indeed, when investigating markers of proliferation, we see that over the time course, PCNA declines until day 17, after which the expression starts to return (Extended Data Fig. 4c). p21^Cip1 follows an inverse pattern with stain intensity increasing initially and then declining slightly by day 31 (Extended Data Fig. 4D). We also saw a decrease in DAPI intensity for days 10 and 17, indicating senescence, but a reversion to control level by day 31 (Extended Data Fig. 4e). To confirm that the predictor accurately determined senescence even 31 days after IR, we evaluated if markers of proliferation and senescence correlated with predicted senescence. Accordingly, cells with predicted senescence had higher p21^Cip1 levels, lower PCNA and lower DAPI intensities and vice versa (Extended Data Fig. 4f–h). Morphologically, area and aspect are higher for predicted senescence, whereas convexity is lower (Extended Data Fig. 4i–k). Finally, a simple nuclei count confirms growth, following IR treatment (Extended Data Fig. 4l). Overall, the senescence predictor captures the state during development in agreement with multiple markers and morphological signs.

Senescent cells are associated with the appearance of persistent nuclear foci of the DNA damage markers γH2AX and 53BP1 (refs. ^31,32). Our base data set including control, RS and IR lines were examined for damage foci using high-content microscopy, where we found the mean count for controls to be below 1 for each marker, whereas RS had 4.0 γH2AX and 2.0 53BP1 foci and IR had 3.4 γH2AX and 3.0 53BP1 foci (Fig. 4a, b and Extended Data Fig. 5a). We calculated the Pearson correlation between predicted senescence and γH2AX and 53BP1 foci counts and found that across all conditions, there is a moderately strong correlation of around 0.5 (Fig. 4c). This association is also visible when simply plotting foci counts and senescence prediction, which shows predicted senescence flipping from low to high, along with shifts in foci counts (Extended Data Fig. 5b). Our feature reduction masked internal nuclear structure, but it is nonetheless notable that senescence prediction correlates with foci count. We also compared the correlation between predicted senescence and area, where we see a correlation of around 0.5. In sum, there is a considerable correlation between foci counts and senescence.

Engineers have designed and successfully tested a more efficient wind sensor for use on drones, balloons and other autonomous aircraft.

These wind sensors—called anemometers—are used to monitor wind speed and direction. As demand for autonomous aircraft increases, better wind sensors are needed to make it easier for these vehicles to both sense weather changes and perform safer take-offs and landings, according to researchers.

Such enhancements could improve how people use their local airspace, whether it be through drones delivering packages or passengers one day flying on unmanned aircraft, said Marcelo Dapino, co-author of the study and a professor in mechanical and aerospace engineering at The Ohio State University.

QUT robotics researchers working with Ford Motor Company have found a way to tell an autonomous vehicle which cameras to use when navigating.

Professor Michael Milford, Joint Director of the QUT Center for Robotics and Australian Research Council Laureate Fellow and senior author, said the research comes from a project looking at how cameras and LIDAR sensors, commonly used in autonomous vehicles, can better understand the world around them.

“The key idea here is to learn which cameras to use at different locations in the world, based on previous experience at that location,” Professor Milford said.