Why ‘open source’ AIs could be anything but, the derailment risks of long freight trains, and breeding better wheat
Jun 26, 2024
auto_awesome
Critics question the 'openness' of AI systems, longer freight trains have higher derailment risk, historic wheat genes offer new traits for current crops
Open source AI models may not meet full openness criteria, impacting research accountability.
Increasing freight train length raises derailment risk, prompting safety concerns and potential regulations in the industry.
Deep dives
AI Models and Openness in AI Research
In the podcast, they discussed the concept of 'open washing' in the realm of AI, where some models may be labeled as open but do not fully meet the criteria for openness. Researchers analyzed the openness of different large language models and found discrepancies between claimed openness and actual criteria fulfillment. The importance of open models is highlighted as the EU's AI act considers requirements based on openness, emphasizing the need for a clear definition of openness in AI research.
Freight Train Length and Derailment Risk
Another topic covered in the podcast explored how longer freight trains could increase the likelihood of derailments. Research based on real-life data showed that as train lengths increased, the risk of derailment also rose. Factors such as weight distribution and safety issues were attributed to the higher risk of longer trains derailing, sparking discussions around potential safety regulations in the railway industry.
Reviving Traits in Wheat for Agricultural Improvement
A significant segment of the podcast delved into wheat research aimed at enhancing the crop's traits by revisiting historic wheat varieties. By crossbreeding old land races with modern wheat types, researchers identified potentially beneficial traits, such as resistance to fungal diseases and nitrogen use efficiency. This extensive breeding effort involved sequencing genomes, creating diverse wheat populations, and analyzing genetic data to understand the desirable traits for future agricultural advancements.
Many of the large language models powering AI systems are described as ‘open source’ but critics say this is a misnomer, with restricted access to code and training data preventing researchers from probing how these systems work. While the definition of open source in AI models is yet to be agreed, advocates say that ‘full’ openness is crucial in efforts to make AI accountable. New research has ranked the openness of different systems, showing that despite claims of ‘openness’ many companies still don’t disclose a lot of key information.
06:12 Why longer freight trains are more prone to derailment
In the US, there are no federal limits on the length of a freight train, but as companies look to run longer locomotives, questions arise about whether they are at greater risk of derailment. To find out, a team analysed data on accidents to predict the chances of longer trains coming off the tracks. They showed that replacing two 50-car freight trains with one 100-car train raises the odds of derailment by 11%, with the chances increasing the longer a train gets. While derailments are uncommon, this could change as economic pressures lead the freight industry to experiment with ever-longer trains.
11:44 How historic wheat could give new traits to current crops
Genes from century-old wheat varieties could be used to breed useful traits into modern crops, helping them become more disease tolerant and reducing their need for fertiliser. Researchers sequenced the genomes of hundreds of historic varieties of wheat held in a seed collection from the 1920s and 30s, revealing a huge amount of genetic diversity unseen in modern crops. Plant breeding enabled the team to identify some of the areas of the plants’ genomes responsible for traits such as nutritional content and stress tolerance. It’s hoped that in the long term this knowledge could be used to improve modern varieties of wheat.