✨98. 🤩 Apple's ChatGPT competitor | 💨 Fervo's geothermal energy testing | 🤝 Tesla in early talks to license FSD to 'major' automaker
PLUS: ChatGPT's decline✨Meta's free AI now available ✨extreme heat's impact on the grid and renewables✨Gulf of Mexicos' first offshore wind lease✨Tesla FSD can transfer to new car for 2 months✨China's SAIC partners with Audi on EVs
⚠️ Today is (probably) the last regular issue of Future Perfect for the immediate future. Henceforth, I want to focus on my law practice’s new AI blog; the first article should (hopefully) go live on Monday.
While I would like to find a way for both Future Perfect and my law practice’s blog to coexist, for now, please bookmark my website; if you remain subscribed to Future Perfect, you’ll continue to get future issues when I’m ready.
Speaking of my law practice, don’t hesitate to reach out if you’re in need of legal or consulting services at the intersection of AI, law, and copyright.
Thanks very much, and have a great weekend!
🤖 ARTIFICIAL INTELLIGENCE
A Stanford University study found that the high-profile AI chatbot ChatGPT, created by OpenAI, performed worse on certain tasks in June compared to its March version.
The study examined ChatGPT's performance over several months in tasks such as solving math problems, answering sensitive questions, generating software code, and visual reasoning.
Significant fluctuations, known as drift, were observed in the chatbot's ability to perform these tasks. For instance, GPT-4's accuracy in identifying a prime number dropped from 97.6% in March to 2.4% in June, while GPT-3.5 showed improved performance in the same task.
The study revealed that changes made to improve the chatbot's performance in specific tasks can have unintended consequences and negatively affect its performance in other tasks.
The reasons for these unpredictable effects are poorly understood, partly due to the lack of transparency in the model's architecture and training data.
In addition to reduced accuracy, ChatGPT also stopped providing step-by-step explanations of its reasoning and avoided engaging with sensitive questions, which decreased transparency and rationale in its responses.
Meta has announced that it is open-sourcing its large language model LLaMA 2, making it freely available for commercial and research use.
This move puts Meta in direct competition with OpenAI's GPT-4, which powers tools like ChatGPT and Microsoft Bing.
The decision to open up LLaMA 2 is aimed at giving businesses, startups, and researchers access to more AI tools and fostering community experimentation.
LLaMA 2 was trained on a larger dataset compared to its predecessor and Meta claims it outperforms other language models in reasoning, coding, proficiency, and knowledge tests.
Meta has emphasized the importance of safety and transparency, stating that LLaMA 2 has undergone rigorous testing and evaluation.
The open-sourced LLaMA 2 will be available through Microsoft's Azure platform, as well as other providers like AWS and Hugging Face.
Apple is reportedly developing its own AI chatbot internally referred to as "Apple GPT" to compete with OpenAI, Google, and others.
The chatbot is built using Apple's own framework called "Ajax" and runs on Google Cloud using Google JAX.
The internal rollout of the chatbot was initially delayed due to security concerns but has since been made available to more Apple employees with special approval.
The chatbot is currently being used for product prototyping, text summarization, and answering questions based on its training data.
Apple is actively seeking generative AI talent and has posted job openings for experts in large language models and generative AI.
Apple aims to address privacy concerns associated with AI and plans to make a significant AI-related announcement next year.
⚡️ POWER PROMPTING!
I'm facing a challenge in [area of challenge]. I'm seeking creative solutions to [describe the specific dilemma or problem]. I'm looking for practical strategies and actionable steps to overcome this challenge. How can I [action or outcome desired] in a way that aligns with my values and goals? Please provide me with customized advice, considering my unique circumstances and any relevant details I should take into account.
Want to submit your own? Tweet me @MarcHoag!
🤖🔥🤯 COOL AI TOOLS, APPS, VIDEOS, PODCASTS, LINKS, AND MORE!
🌡️ CLIMATE CHANGE & CLEAN ENERGY
Fervo Energy has successfully completed a well test at its Project Red geothermal site in northern Nevada, confirming the commercial viability of its drilling technology.
The 30-day well test achieved a flow rate of 63 liters per second, enabling 3.5 MW of electric production, setting new records for flow and power output from an enhanced geothermal system.
Fervo is the first company to drill a horizontal well pair for commercial geothermal production, utilizing drilling techniques developed by the oil and gas industry.
The success of the test supports Fervo's claim that geothermal energy could supply over 20% of US power needs and shows there are no technological barriers to geothermal deployment.
Fervo aims to produce 400 MW by 2028 and is working on greenfield developments in Utah to scale up its geothermal projects.
The use of fiber optic sensing tools and the implementation of seismic activity mitigation protocols have been key factors in Fervo's success in accessing geothermal resources and reducing costs.
The US Department of the Interior will hold the first-ever offshore wind energy lease sale in the Gulf of Mexico.
The lease sale will include three areas off the coast of Lake Charles, Louisiana, and Galveston, Texas, with the potential to generate 3.7 GW of power.
The auction is scheduled for August 29, 2023, and aims to advance the development of offshore wind energy in the Gulf of Mexico.
The Biden administration has a goal of deploying 30 GW of offshore wind by 2030 and achieving net-zero electricity in the US by 2035.
This lease sale marks another step towards creating a clean energy future, tackling the climate crisis, and creating jobs, according to Secretary of the Interior Deb Haaland.
Extreme heat can strain the power grid and reduce the efficiency of power generation and transmission, potentially leading to blackouts.
Solar panels can experience reduced output efficiency of 10-25% in extreme heat due to declining voltage.
High temperatures can coincide with calm weather conditions, reducing wind power generation.
Excessive heat can accelerate battery degradation and increase the risk of thermal runaway in lithium-ion battery storage systems.
Innovations such as thermophotovoltaics, advanced solar panel designs, predictive weather models, and improved battery thermal management systems are being explored to counteract the negative effects of extreme heat on renewables and battery storage.
Incorporating a mix of renewables and energy storage systems can help create a more resilient grid during extreme heat events.
🚗 AUTONOMOUS & ELECTRIC VEHICLES
Tesla will allow owners who purchased the Full Self-Driving (FSD) package to transfer it to a new car, but only if they place an order this quarter.
Tesla has been selling the FSD package since 2016 with the promise of achieving self-driving capability, but has yet to deliver on that promise.
The FSD package has increased in price over the years, causing frustration for owners who paid for the package but haven't received the promised features.
Tesla's offer to transfer the FSD package to a new car is described as a "one-time offer of amnesty" and is only valid for orders placed before the end of September.
While the offer is appreciated, some feel that it should be available to all owners who purchased the FSD package since 2016, as Tesla has not delivered on its promises.
This offer allows Tesla owners with older vehicles to upgrade and gives Tesla another opportunity to deliver on the FSD features.
Elon Musk confirmed that Tesla is in early discussions to license its self-driving technology to a major automaker.
Integrating Tesla's self-driving technology into other automakers' vehicles would require integrating Tesla's self-driving computer and cameras.
Tesla's camera-based strategy differs from most other automakers who rely on different sensors like lidar.
Other automakers often have separate purpose-built vehicles for autonomy, while Tesla aims to achieve level 4 or 5 autonomy through software updates.
Tesla's increased communication channels with other automakers, including Supercharger partnerships, may have facilitated these discussions.
It remains to be seen which automaker Tesla is in talks with and if a partnership will be formed.
SAIC Motor, a state-owned automaker in China, has confirmed a partnership with Audi regarding next-generation EV platforms.
Audi had been exploring the Chinese market for a new partner to acquire advanced EV platform technology following internal software issues and development setbacks at Volkswagen Group.
SAIC Motor is expected to provide technology for platforms powering EVs from its luxury sub-brand IM Motors.
The partnership aims to accelerate EV development and deepen strategic cooperation between SAIC Motor and Audi.
The extent of the partnership and whether Audi will acquire the platform from IM Motors or co-develop a new architecture with SAIC Motor remains unclear.
Further details and comments from Audi are awaited.
🎉 THAT’S ALL FOR TODAY!
What'd you think of today's email?
This helps me improve it for you!
✅ 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!
Looking for past newsletters? You can find them all here.