Breaking the Bottleneck | Issue 36
[1/8/2023] 2024 Predictions, DeepMind Robotics, EV Manufacturing, Tulip, & Waymo's Safety
Breaking the Bottleneck is a weekly manufacturing technology newsletter with perspectives, interviews, news, funding announcements, and a startup database.
Happy New Year! I will be posting the next version of my discrete and continuous manufacturing market map next week. If you have any suggestions or know of any companies that might be a fit I’d love to chat!
Content I Enjoyed This Week 🏭🗞️🔬
News:
A Shortlist of 2024 Manufacturing Predictions
2024 Manufacturing Industry Outlook [Deloitte]
2024 Aerospace & Defense Outlook [Deloitte]
2024 Chemicals Outlook [Deloitte]
JPM2024: Manufacturing Outlook for Biopharma [Hogan Lovells]
Industrial Cybersecurity Trends Part 1 & Part 2 [Manufacturing.net]
Additive Manufacturing in 2024 [Voxel Matters]
3D Printing Predictions [3D Print]
Shaping the Future of Advanced Robotics [Google DeepMind]
Google's DeepMind Robotics researchers are among the teams exploring the potential of the convergence of generative AI and large foundational models with robotics. They have introduced two key innovations:
AutoRT: This system utilizes large foundational models to improve robots' understanding of human desires. It uses a Visual Language Model (VLM) for better situational awareness. AutoRT can coordinate a group of robots equipped with cameras to map their environment and objects within it. Large language models suggest tasks for the robots to perform, reducing the need for explicit programming.
RT-Trajectory: This innovation uses video input for robot learning. It incorporates a two-dimensional arm sketch over the video to aid the robot in understanding its control policies. DeepMind reported a 63% success rate with RT-Trajectory, compared to 29% with previous methods, during testing of 41 tasks. This approach leverages underutilized information in existing robot datasets.
Intel Is Building Gen AI To Improve Yield [SemiAnalysis]
Intel is developing a deep generative model called GenAI for predicting device variation in chip manufacturing. They are using generative adversarial networks (GAN) and diffusion models for this purpose. While GANs are commonly used for image, text, and audio generation, diffusion models are better suited for predicting chip manufacturing process yield as they can replicate the long tails of the sample data distribution. In Intel's research, they use SPICE parameters as input for the deep learning model and predict electrical characteristics of the device as manufactured (ETEST metrics). This helps optimize chip yields at the design stage, reducing costs and development times. While the work is still in the research stage, it's expected that major foundries and design firms will work to industrialize similar techniques.
Machine Learning Helps Fuzzing Find Hardware Bugs [IEEE Spectrum]
Researchers at Texas A&M University are using a technique called "fuzzing" to automate chip tests on the assembly line and discover hardware vulnerabilities. Fuzzing involves introducing commands and prompts to a chip that are not quite correct, causing the system to behave unpredictably. This helps researchers identify potential weak links in the system that hackers could exploit. The researchers used reinforcement learning to select inputs for fuzz testing, achieving significant speedup in detecting vulnerabilities. Automated hardware-testing techniques like fuzz testing are a first line of defense to uncover easy-to-find vulnerabilities, freeing up security experts' time to focus on more complex bugs.
India Wanted a Manufacturing Boom, Workers Are on the Farm Instead [WSJ]
The lockdown in 2020 led to an exodus of workers from cities back to rural areas in India. Even as the pandemic receded, India's farm sector added 13 million workers in the past year, all driven by Narendra Modi's extension of a critical food program for another five years. Consequently, jobs in manufacturing have remained stagnant, and factories are struggling to hire. This trend represents a premature form of deindustrialization in India, contrary to what many economists anticipated. Instead of a transition to factory work, India's farm workforce has grown significantly, posing risks to the country's utilization of its large labor force and economic growth potential. Even with all the talk around India as Asia’s next manufacturing hub, manufacturing's contribution to India's GDP has declined from 17% two decades ago to 13% in 2022, with only five million new manufacturing jobs added since Modi's first term. India's economic growth seems more aligned with worker preferences for working in less labor-intensive sectors like information technology and financial services.
Battery Belt Communities Fight for Benefits [Canary Media]
"BlueOval City" the automotive manufacturing complex and a joint venture between Ford and SK Innovation, promises economic growth and job opportunities in rural communities. However, many residents in these communities have concerns about the large-scale development's impact on their way of life, natural resources, and potential influx of newcomers. To address these concerns, a coalition of local organizations and residents is negotiating a community benefits agreement (CBA) with Ford and SK Innovation. This CBA outlines various stipulations and demands, including community resources like youth facilities, support for road maintenance, apprenticeship programs, waste disposal assurances, and input in community programs. CBAs have historically been used in the entertainment and sports industries but are now increasingly employed in clean energy projects, with several agreements signed in recent years. These agreements aim to give local communities a say in projects that impact them, ensuring accountability, transparency, and local input.
Research:
EDF Analysis on U.S. Electric Vehicle Battery Manufacturing [EDF]
An analysis by the Environmental Defense Fund (EDF) reveals that the United States already has enough battery production capacity in the pipeline to supply all the expected electric vehicles (EVs), including cars and trucks, to be sold in 2030. More than 1,000 gigawatt hours per year of U.S. EV battery production capacity has been announced to come online by 2028. This capacity is equivalent to powering 10 million electric cars, which is more than what the U.S. Environmental Protection Agency (EPA) projects to be sold in 2030. The analysis highlights the rapid growth of the EV market, with electric vehicles reaching 12% of all vehicles produced for sale in the U.S. in 2023, a 70% increase from the previous year. All the announced battery manufacturing is domestic, emphasizing the U.S.'s self-sufficiency in battery production. Key states with announced battery production capacity include Michigan, Georgia, Tennessee, Kentucky, and Indiana.
Scope 3 Emissions Categories in the Automotive Supply Chain [Bain]
Metal AM Report [Voxel Matters]
Currently valued at around $2.8 billion, the metal AM market is a small fraction of the overall metal manufacturing market. However, it is expected to grow substantially, to $40.3 billion within the next decade. Today, the industry is transitioning from small-batch to medium-batch and large-batch production of final parts, promising scalability, efficiency, and environmental sustainability. Additionally, companies are focusing on developing and optimizing metal alloys for industrial use cases, leading to a rapid expansion of available materials across aerospace, automotive, medical, maritime, and energy.
Podcasts:
A New Era of Design Engineering [Manufacturing Executive]
Unlocking the Full Potential of Frontline Operations, Lean Manufacturing, & I4.0 w/ Natan Linder of Tulip [TBD]
Manufacturing Deals
GreyOrange - A company providing retailers, warehouse operators and third-party logistics providers (3PLs) around the world with automated robotic fulfillment and inventory optimization solutions
$135 million [Series D] - Led by Anthelion Capital and joined by existing investors Mithril, 3State Ventures, & Blume Ventures
Weekly Planned Downtime
Happy New Year!