NextBigFuture.com

The distilled versions of Deepseek are not as good as the full model. They are vastly inferior and other models out perform them handily. Running the full model, with a 16K or greater context window, is possible for about $2000 at about 4 tokens per second. This uses an Machine Specs AMD EPYC 7702 512GB ... Read more

What if Tesla succeeds with robotaxi in June 2025 in Austin? Will there be a buying frenzy for Tesla Robotaxi?

Tesla has released a video showing robotic cleaning of the Tesla Cybercab. This will greatly improve the economics of robotaxi. Farzad and others have estimated cleaning as a potentially big cost of Tesla Robotaxi operations. If cleaning with robots drops the cost of robotaxi operations to $100 per month for 3400 miles of operation then ... Read more

Some believe DeepSeek is so efficient that we don't need more compute and everything has now massive overcapacity because of the model changes. Jevons Paradox is closer to reality because demand has already increased H100 and H200 pricing. Deepseek and High Flyer have a mix of 50,000 H20s, H800s, A100s and H100 GPUs. Deepseek has ... Read more

META CEO Mark Zuckerberg predicts that 2025 is the year that an AI assistant will serve over one billion users and he thinks META will be company that provides that AI assistant. In 2023, Meta revealed they have AI inference accelerators that they are designing in-house specifically for Meta's AI workloads. Deep learning recommendation models ... Read more

San Francisco, United States / California, 30th January 2025, CyberNewsWire

Palo Alto, USA, 30th January 2025, CyberNewsWire

Tesla has announced they they will have paid robotaxi without human drivers in Austin in June 2025 and they will thousands and perhaps ten thousand Optimus teslabots this year. Third megapack factory is already being built but they did not say where yet. Tesla will also deploy robotaxi to California and other locations in the ... Read more

Tesla has its 2024 update on financials and operations. Plans for new vehicles, including more affordable models, remain on track for start of production in the first half of 2025. These vehicles will utilize aspects of the next generation platform as well as aspects of our current platforms and will be produced on the same ... Read more

NuScale and Standard Power are developing two new nuclear reactors facilities in Ohio and Pennsylvania powered by its SMR technology. The plan is to build 24 units of its 77 MWe modules, producing 1,848 MWe energy from both sites. Standard Power thinks this facility could be operational by 2029 at the earliest. Nuscale has developed ... Read more

SpaceX Starlink and Apple have partnered to bring Starlink direct to cellphone connections to iPhone. Starlink works right from your pocket. T-Mobile to beta testers: You can now stay connected with texting via satellite from virtually anywhere" While it's text-only for now, SpaceX and T-Mobile plan to add voice calls and data in the future. ... Read more

Tesla cars will now have their first drives from the Factory to the loading dock without human drivers. Teslas now drive themselves from their birthplace at the factory to their designated loading dock lanes without human intervention One step closer to large-scale unsupervised FSD pic.twitter.com/Aj6dHsLaRO - Tesla AI (@Tesla_AI) January 29, 2025 Barry Glack says ... Read more

Jim Fan tells AI community don't worry. Be happy...and grind on code 24-7. Faster is the only way. Many tech folks are panicking about how much DeepSeek is able to show with so little compute budget. I [Jim Fan] see it differently - with a huge smile on my face. Why are we not happy ... Read more

There are many big questions about the impact of the efficiency gains from the improved DeepSeek methods. Deepseek improved the Reinforcement learning. Deepseek also directly accessed the Nvidia chips. They did NOT use Nvidia CUDA. CUDA was preventing them from doing what they needed to do with the chips. The Deepseek research paper gave a ... Read more

The Opensource DeepSeek R1 model and the distilled local versions are shaking up the AI community. The Deepseek models are the best performing open source models and are highly useful as agents and for other tasks. Here are videos that guide people through using the models. There will soon be many options for Deepseek variants ... Read more

Altimeter Capital analyst and partner puts what Deepseek claims and results into numbers. $6M Training Costs = Plausible IMO Quick math: Training costs (active params * tokens). DeepSeek v3 (37B params; 14.8T tokens) vs. Llama3.1 (405B params; 15T tokens) = v3 theoretically should be 9% of Llama3.1's cost. And the disclosed actual figures aligned ... Read more

Langchain used ollama to install Deepseek 14B on a laptop. They used for a local deep researching model. $ ollama pull deepseek-r1:14b $ export TAVILY_API_KEY= $ uvx -refresh -from langgraph-cli[inmem]" -with-editable . -python 3.11 langgraph dev

Making AI to 10 to 30 times more efficient for AI inference and getting more value from training will increase AI demand. Increased AI efficiency on training and inference will accelerate the improvement and usefulness of AI. Dr Know It All went over the DeepSeek paper and explains how they automated the Reinforcement Learning. AlphaZero ... Read more

DeepSeek just dropped a new open-source multmodal AI model, Janus-Pro-7B. It is MIT opensource license. It's multimodal (can generate images) and beats OpenAI's DALL-E 3 and Stable Diffusion across GenEval and DPG-Bench benchmarks. This comes on top of all the R1 hype. Here is the link to the Deepseek Janus 7B Github. NEWS: DeepSeek just ... Read more

Youtuber Ominous Industries provides a training video for setting up your Jetson Nano, installing Pytorch and setting up your own (likely open source LLM). This takes the NVIDIA Jetson Nano to the next level by training a Large Language Model (LLM) completely from scratch. It uses the lightweight and efficient NanoGPT framework, This covers setting ... Read more

Cary, North Carolina, 26th January 2025, CyberNewsWire

Youtuber, Ominous Industries, ran a couple of versions of the DeepSeek R1 1.5B of models running locally on the NVIDIA Jetson Nan. The newly released distilled DeepSeek models were explroed. The DeepSeek R1 1.5B model delivers impressive performance with plenty of room to spare on the Jetson. He shows the installation process, followed by a ... Read more

It is possible to load and run 14 Billion parameter llm AI models on Raspberry Pi5 with 16 GB of memory ($120). However, they can be slow with about 0.6 tokens per second. A 13 billion parameter model can run at 1.36 tokens per second. Improved firmware (better SDRAM timing) improved results.

In March ,2024, Nextbigfuture started covering NANO Nuclear. It was then a startup that had raised over $8 million to develop micro nuclear fission reactors with up to 2 megawatts of power. The reactors will be transported by Semi Trucks. I, Brian Wang of Nextbigfuture, was contacted by the Nano Nuclear team for a correction ... Read more

The Opensource Deepseek R1 AI model is top notch in terms of math, reasoning and coding but it has a lot of bias and restrictions. Here is a video describing some of the examples. It will describe how to use VPNs to get around the China firewall. It provides a Chinese communist part version of ... Read more

Deepseek is freely available online for simple chat interface and charges far lower costs for heavy million+ token usage. The huge increase in capability with a team of just over 100 developers is causing panic at Meta and other AI companies. All AI competitors will have to rethink and rework what they are doing. Everyone ... Read more

SpaceX Starlink direct from satellite to cell phone Internet connection starts beta test in 3 days. Starlink direct from satellite to cell phone Internet connection starts beta test in 3 days https://t.co/ygAjtTN8SY - Elon Musk (@elonmusk) January 24, 2025 T-Mobile is offering Starlink texting to cellphones, but it's currently in a limited beta phase: T-Mobile ... Read more

Demis Hassabis is the CEO of Google DeepMind. Demis thinks new agent type systems will be able to perform additional tree of knowledge search to achieve breakthrough insights like the move 37 in the Go system game. In this conversation, he talks about the path to artificial general intelligence. How long it will take to ... Read more

Experimenters have had overnight tests confirming they have OPEN SOURCE DeepSeek R1 running at 200 tokens per second on a NON-INTERNET connected Raspberry Pi. This is a distilled smaller model than the OPenAI O1 class model. Folks, I think we have done it! If overnight tests are confirmed we have OPEN SOURCE DeepSeek R1 running ... Read more

There has been the controversial announcement what is called a $500 billion AI project in the USA. There is controversy about how much funding is in the project. The announcement did talk about building this out over years. Altimeter Capital has an analysis of the financials for the $100 billion per year of the Softbank, ... Read more

Angry Astronaut and Nextbigfuture commenter are making the case that SpaceX and Elon Musk must switch to nuclear thermal rockets to colonize Mars. I will review that the nuclear thermal rocket program is taking and will take far longer. Also, triple the ISP does not cut the travel time by one third. There is no ... Read more

The fundamental change for Tesla needs to happen in 2025 to transform financials of the company. The Nvidia Chatgpt moment led to a surge of over 10X in Nvidia's share price. There was also a surge in revenues, earnings and margin. This is the potential for Tesla if all of the growing AI related capabilities ... Read more

xAI Grok 3 will out in about two weeks. It has been previewed to Dr Alan Thompson. It is good at writing code and modelling the real world. It made code in python for ball bouncing inside a square and inside a tesseract. I project it is possible that xAI will release more major models ... Read more

There is report that Tesla has given suppliers guidance on Optimus production: 600 units per week by the end of 2025. Ten weeks at the rate of 600 Optimus bots per week would by 6000 units. Tesla has given suppliers guidance on Optimus production: 600 units per week by the end of 2025. Wow, what ... Read more

Torrance, United States / California, 22nd January 2025, CyberNewsWire

In September, Verseon, a pioneering company in physics- and AI-powered drug discovery, revealed that its AI technology significantly outperforms Google's cutting-edge Deep Learning models in terms of prediction accuracy for a wide range of datasets. Just a month later, the company announced a breakthrough in combining multiple AI models in dynamic ways. And most recently, ... Read more

On January 21, 2025, President Trump, Softbank, OpenAI, Oracle and MGX announced details of the Stargate project, with an initial commitment of $100 billion and plans to invest up to $500 billion over the next four years. Oracle's Larry Ellison, SoftBank CEO Masayoshi Son, and OpenAI Sam Altman were present to jointly announce the formation ... Read more

100 person China startup DeepSeek just released the first Open Source Reasoning Model that matched the OpenAI o1 reasoning model. OpenAI was charging $200 per month to use the OpenAI o1 pro model. How did an unknown, 100 person startup with $0 VC funding produce a frontier open source model that rivaled OpenAI and Anthropic ... Read more

DeepSeek R1 is an open sourced model. DeepSeek is a Chinese AI research company backed by High-Flyer Capital Management, a quant hedge fund focused on AI applications for trading decisions. They have released models under open-source licenses like MIT. How did they match or even surpassing OpenAI's O1: Reinforcement Learning Focus: DeepSeek-R1 and its variant, ... Read more

Lithium-sulfur battery and supermaterials firm Lyten is seeking a US$650 million loan from the US import-export bank EXIM to scale up manufacturing and meet BESS orders from the Caribbean region. Lyten has received letters of interest (LI) from the Export-Import Bank of the United States (EXIM) in support of the financing package, it said this ... Read more

Safely and prudently sending a manned mission that arrives and lands before Trump leaves office is possible. Every 26 months there are launch windows to send missions from Earth to Mars. SpaceX is working toward sending five SpaceX Starships to Mars in 2026. The unmanned missions in 2026 could expand with government support. President Trump ... Read more

The US in creating an external revenue service to collect tariffs. This is re-opening negotiations for all trade and other relationships with all other countries. It is the threat of actions and actual actions where the economic strength of the United States will be used to pressure other countries.

Over the last few years, artificial intelligence has become the biggest talking point in tech. But conversations centered around what AI can achieve - now and in the future - overlook an important point. Namely, that the entry-point to AI must be simple and seamless in order for widespread adoption to occur. Above - Source: ... Read more

Trump said We will pursue our manifest destiny into the stars by launching American astronauts to plant the Stars and Stripes on the planet Mars". Elon Musk's reaction to Trump saying today: We will pursue our manifest destiny into the stars by launching American astronauts to plant the Stars and Stripes on the planet Mars." ... Read more

Researchers at Oregon Health & Science University have uncovered how a molecule found on certain bacteria may drive blood clotting in sepsis, a life-threatening condition that causes about 8 million deaths per year. The team in the cardiovascular engineering lab at OHSU has focused on the role of specific blood clotting mechanisms in sepsis, with ... Read more

Sam Altman, the CEO of OpenAI, now says he is more confident of a fast AI takeoff. A fast AI takeoff is more possible than he thought a couple of years ago. He suggested this could happen within a small number of years rather than a decade. This indicates a shift in his perspective towards ... Read more

Falcon Heavy by SpaceX has a cost per kilogram to LEO of approximately $1,400 per kg. This figure reflects the cost-effectiveness achieved through partial reusability and high payload capacity. A single use Super Heavy Starship and booster will be able to bring full payloads to orbit for about $250-600 per kilogram. This is with costs ... Read more

OpenaI internal news is that they have progress to AI that can innovate. UPDATE: Sam Altman says there no AGI in the next month and no AGI has been built. twitter hype is out of control again. we are not gonna deploy AGI next month, nor have we built it. we have some very cool ... Read more

The catch of SpaceX Starship flight 7 had vastly improved efficiency. Look at how much cleaner and perfect the Flight 7 catch was than the Flight 5 one! Congrats @spacex! pic.twitter.com/YlEVyHp4Ue - Starship Alves (@StarshipAlves) January 18, 2025 Catches: Starship 5 vs Starship 7 What an amazing time to be alive. Such an exciting ... Read more