- Future Perfect
- Posts
- ✨95. 🤪 AI goes crazy when trained on AI-generated data | ⚡️ bifacial solar cells | ⚖️ AV software IP consierations
✨95. 🤪 AI goes crazy when trained on AI-generated data | ⚡️ bifacial solar cells | ⚖️ AV software IP consierations
PLUS: EU AI Act needs developers at the table✨China's blueprint for controlling AI✨5 GW solar panel order ✨US heatwave✨Blue Bird EV school buses✨Cruise claims humans are bad drivers
🤖 ARTIFICIAL INTELLIGENCE
Policymakers in the EU are finalizing the A.I. Act, the world's first comprehensive regulation for artificial intelligence.
The A.I. Act aims to encourage responsible A.I. innovation while mitigating risks such as bias and misinformation.
The Act introduces risk-based requirements for A.I. systems in various sectors, with high-risk areas including critical infrastructure and access to education and employment opportunities.
The legislation recognizes the importance of open-source development and includes provisions to protect open-source developers.
Other countries, including the United States and Canada, are considering or proposing their own A.I. regulations, with a focus on a risk-based approach and support for open-source collaboration.
Feeding AI-generated content to AI models causes a decline in output quality and can break the model's digital brain, according to scientists at Rice and Stanford University.
The process of training AI models repeatedly on synthetic content leads to a condition called Model Autophagy Disorder (MAD), where the model's precision and diversity progressively decrease.
The reliance on synthetic data and the absence of fresh real data from human work can negatively impact the outputs of AI models.
The findings have implications for AI models trained on scraped online data, as the prevalence of synthetic content on the internet could affect the quality and structure of the open web.
Adjusting model weights and considering the need for human input could help mitigate the potential negative effects of AI models going MAD.
The study raises questions about the usefulness of AI systems without human involvement and highlights the limitations of AI in replacing human creativity and input.
China has released new regulations for generative AI, which will be overseen by the country's top internet regulator, the Cyberspace Administration of China (CAC).
The regulations apply to publicly available chatbots and address issues such as intellectual property rights, privacy protection, and social mobilization ability.
Chinese tech companies have been awaiting these regulations before releasing their generative AI tools for public use.
Developers will need to register their algorithms with the Chinese government and undergo a security assessment, particularly if their services can influence public opinion.
The regulations offer insights into China's global ambitions for AI and may serve as a guide for other countries in dealing with copyright infringement, data protection, and international cooperation in AI regulation.
⚡️ POWER PROMPTING!
Imagine you were my personal life coach. Based on the key values I've described to yHere's how you could use Python to tackle this issue:
How could you use natural language processing (NLP) to categorize these emails into different types of requests or complaints?
What machine learning algorithms might be useful to rank emails by their urgency, based on keywords or phrases?
How could you automate the process of sending acknowledgment emails or even simple canned responses, depending on the categorization?
How might you use data visualization libraries in Python to present the categorization and urgency data in a digestible format for the team?
Please describe your approach to writing this Python script, including libraries you might use, your overall strategy for both categorization and prioritization, and how you'll handle potential challenges (e.g., recognizing sarcasm or other nuances of human language).
Want to submit your own? Tweet me @MarcHoag!
🤖🔥🤯 COOL AI TOOLS, APPS, VIDEOS, PODCASTS, LINKS, AND MORE!
CoPilot2Trip.com: AI-powered trip planning with maps.
StockIMG.ai/text-styles: AI-powered decorative text generation.
Albus Chrome Extension: ChatGPT-4 powered Chrome extension for enhanced browsing capabilities.
🌡️ CLIMATE CHANGE & CLEAN ENERGY
First Solar, the largest solar panel maker in the US, has secured a 5 GW order from Israeli renewable energy company Energix.
The ultra-low carbon thin film solar panels will be delivered to Energix between 2026 and 2030 for projects in Israel, Poland, and the US.
Energix will also collaborate with First Solar Recycling Services for managing end-of-life solar panels through a recycling program.
With this order, Energix will operate between 7 and 9 GW of First Solar panels by 2030.
First Solar's success in securing large orders is attributed to the Biden administration's Inflation Reduction Act, which incentivizes US-based renewable manufacturing.
First Solar is investing in expanding its US manufacturing capacity and developing its cadmium telluride (CdTe) PV technology, known for its low carbon and water footprint.
The US is experiencing extreme heat, with temperature records being broken across the country.
The Southwest and East Coast are under extreme heat stress.
Over 110 million people, nearly a third of the US population, are under excessive heat warnings and advisories.
A heat dome is intensifying the heatwave, with around 100 heat records expected to be broken.
The heat is affecting various activities, such as concerts being cut short and flight cancellations.
It is crucial to prioritize safety by staying indoors, finding air conditioning, and staying hydrated in these extreme heat conditions.
Bifacial solar cells have the potential to produce higher energy yields at lower costs compared to monofacial cells.
A recent study by the US Department of Energy's NREL found that bifacial perovskite solar cells perform effectively from both sides.
In the study, the front-side efficiency of the bifacial cells exceeded 23%, while the back-side efficiency was about 91-93% of the front.
The ideal thickness of the perovskite layer and rear electrode were determined through simulations to maximize efficiency and minimize resistive loss.
Bifacial perovskite solar modules may have a higher manufacturing cost, but their increased power generation (10-20% more) offsets the upfront expenses.
🚗 AUTONOMOUS & ELECTRIC VEHICLES
Blue Bird, a renowned bus manufacturer, has introduced an improved version of its electric school bus.
The new Vision electric school bus, developed in collaboration with Accelera by Cummins, has 25% more energy storage capacity with a 196 kWh battery and provides 130 miles of range on a single charge.
It has a faster charging rate, up to 80 kW, allowing the battery to fully recharge in three hours.
The increased battery capacity facilitates vehicle-to-grid utilization, allowing excess energy stored in the bus battery to be sold back to electric power companies.
The new electric school bus is lighter, 1000 pounds lighter than the previous generation, and has a larger passenger capacity of 77.
The bus performs better in cold weather conditions and offers an optional fuel-fired heating system for improved cabin heat.
Blue Bird aims to increase electric bus production with its new "EV Build-up Center" capable of producing 5,000 electric school buses per year.
Autonomous vehicle companies claim that their technology will save lives by eliminating human error in driving.
However, the idea that self-driving cars are the clear solution to reducing traffic deaths is deceptive and self-serving.
While autonomous vehicles may eliminate certain mistakes made by human drivers, they can also make their own errors in situations that humans could handle.
The widespread adoption of self-driving technology is still decades away, and hundreds of thousands of lives could be lost in the meantime.
There are immediate strategies that can be implemented to reduce crash deaths, such as redesigning roads, reducing speed limits, and investing in public transportation.
The European Union has seen a decline in roadway deaths without the use of self-driving cars, emphasizing the effectiveness of other measures.
Autonomous vehicles incorporate unique forms of technology requiring protection through intellectual property rights. An effective IP strategy incorporates both patents and copyrights to protect against infringement and monetize the technology.
Patent protection involves overcoming anticipation or obviousness based on prior art, and avoiding classification as a "judicial exception". Utility patents safeguard functional aspects of an autonomous vehicle for 20 years, while design patents protect aesthetic components for 15 years.
Software patents are essential to protect the application of ideas, rather than the ideas or maths alone, often referred to as a "technical effect". This is evident in the protection of specific software programs executing autonomous driving using AI on LIDAR and computer vision inputs.
Copyright protection, lasting for the life of the author plus 70 years, covers the code itself. However, challenges arise because the software must be a work of authorship, original, and fixed in a tangible medium. AI-generated works are currently not recognized as products of human authorship, therefore, are not protected by copyright law.
Collaborations between manufacturers and smaller entities when developing autonomous vehicle software have raised issues about the copyrightability of AI/ML trained models and ownership of the input and output data. Clear definitions of ownership are crucial.
Licensing agreements for AI and machine learning software in autonomous vehicles pose challenges in defining ownership of trained models and derivatives. Comprehensive provisions in licensing agreements are recommended to avoid potential litigation.
🎉 THAT’S ALL FOR TODAY!
What'd you think of today's email?This helps me improve it for you! |
🙏 PLEASE:
✅ Share this post if you liked it
✅ Share this link (https://www.futureperfect.news) if you're a fan of Future Perfect
✅ Vote above before leaving!
Thanks for reading!
-Marc 👋
Looking for past newsletters? You can find them all here.
Reply