Welcome to Binary Circuits’ 15th weekly edition
Your weekly guide to most important developments in technological world
Dear Readers,
Binary Circuit investigates trends, technology, and how organizations might profit from rapid innovation. GreenLight, the brain behind Binary Circuit, finds possibilities, analyzes challenges, and develops plans to fit and grow firms to stay ahead. Looking to scale your business, partners, or AI/technology integration, schedule a free discovery call:
This week, we discuss:
AI Spending: Buy, Don't Build: AI spending is shifting towards commercial solutions. What does it mean for your strategy?
Brain-to-Speech Breakthrough: New BCI tech enables real-time speech from brain signals, reaching up to 91 words per minute — a game-changer for assistive communication.
OpenAI's Big Funding & New Model: Backed by SoftBank, OpenAI ramps up infrastructure and launches an open-weight model, aiming for $12.7B in revenue by 2025.
4 GenAI Truths Leaders Must Face: MIT’s Huttenlocher urges realistic adoption — GenAI needs human judgment, cross-functional teams, and culture change to succeed.
Let’s dive in.
AI spending is poised to rise and shift towards commercial solutions. How does this affect your enterprise strategy?
Gartner notes that the global spending on GenAI is projected to reach $644 billion by 2025, representing a 76.4% increase from 2024. Other research supports this claim; AI at Wharton indicates that spending has grown by 130%. Additionally, Deloitte reports that 74% of companies have met or exceeded their AI goals.
Hardware will take about 80% of all AI spending, and the trend will continue in the foreseeable future. Gartner explains that the shift to software won't happen as many believe. Instead, AI features will be built into existing software at no extra cost. By 2027, it will be almost impossible to buy a PC that is not AI-enabled."
Buy, Don't Build. Companies are now moving away from building their own AI systems and buying ready-made solutions that work better and deliver clearer value. Organizations are moving from custom GenAI systems to commercial options due to difficulties and erratic ROI. Technical executives should give vendor solutions incorporating AI top priority in current systems.
This means making vendor solutions that include GenAI capabilities the top priority for company executives instead of starting from scratch.
Implications for your company. Spending the most money will not bring success with GenAI. Success will result from matching investments with organizational readiness, tackling issues such as data shortages and change resistance, and selecting clearly beneficial solutions.
Brain-computer interfaces ready to match natural conversational speeds
The new brain-to-voice neuroprosthesis converts brain activity into speech in real-time. Tested on a woman with quadriplegia, it transformed communication for people who are severely crippled by attaining almost typical conversation speeds. Additionally affecting voice recognition is the technology that analyzes brain signals in minute increments.
The first brain-to-voice system trained on silent speech attempts to decipher speech intentions and innovative words. Larger vocabulary and surgical implantation reduce accuracy, but research intends to improve expressiveness and spread its use.
Researchers used AI algorithms on the brain-computer interface (BCI) to interpret phrases as the woman thought of them and utter them in a synthetic voice.
Past brain-computer interfaces required 20 seconds or more to translate a thought into speech. This new system reduces lag time to about one second, allowing for natural conversation.
Important statistics:
With a 12.3% mistake rate for popular phrases, the system achieves speeds of up to 90.9 words per minute, delivering almost conversational fluency.
Adaptability to foreign vocabulary exhibited 46% accuracy, therefore displaying strong generalization of speech patterns.
On average, it converts brain signals linked to intended speech at 47 words per minute, almost half the speed of normal conversation in some instances.
How does this technology operate? The implant records signals from speech-related brain areas despite muscle paralysis.
Computers operating an artificial intelligence deep learning system receive signals that translate brain patterns into text and synthesize speech from the imagination of more than 1,000 words.
The experiment especially showed that artificial intelligence could learn from other people's data and operate with signals from different hardware.
For assistive technology, the "streaming speech neuroprosthetic" uses the fast decoding observed in devices such as Alexa and Siri to provide more natural communication for persons who have lost their voice.
OpenAI's open-weight language model and $40 Billion Funding made headlines this week in the AI world
SoftBank's $30 billion investment kicks off OpenAI's historic $40 billion funding round, valuing the business at $300 billion. The money is meant to advance the Stargate project, artificial intelligence infrastructure, and research. Stargate will receive $18 billion, and computational infrastructure will be enhanced
OpenAI's financial release amid legal scrutiny and Elon Musk's lawsuit depends on its for-profit strategy for 2025. However, OpenAI forecasts revenue to rise to $12.7 billion by 2025, reflecting the $1 trillion revenue within a decade.
OpenAI has been concurrently developing its first open-weight language model based on GPT-2 since 2019. This approach facilitates complex ideas, multilingual problem-solving, and customization. Strong credentials include participation in the top 500 USA Math Olympiad, being in the 89th percentile in competitive programming, and demonstrating PhD-level accuracy in physics, biology, and chemistry.
Even with English input, the model can tackle issues in various languages, including Chinese and Persian. Its open-weight structure allows for customization without reliance on the original training data.
The OpenAI philosophy links scale and infrastructure investment to transparency and customization. The $40 billion influx reaffirms the open-weight model's economic dominance and revitalizes its collaboration with intellectual and developmental communities.
Four truths about GenAI that leaders must acknowledge
Dean of MIT's Schwarzman College of Computing and Amazon board member Daniel Huttenlocher states that generative AI is powerful, but it’s not magic. Companies must be realistic, open to experimentation, and reconsider how the technology could complement their staff, products, and culture.
He also enumerated four "inconvenient truths" regarding GenAI that companies sometimes ignore:
Practical agents require more than just language prediction. Although the GenAI models are good at generating text and graphics, today's GenAI is not very good in terms of dependability and challenging task execution.
GenAI lacks human perception. Even the finest models do not reason the way people do. However, in high-stakes industries like healthcare, it performs better when combined with human judgment.
AI change is cultural as much as technological. Teams that want to truly gain from GenAI must be cross-functional. It's about solving real problems in the right context, not just about coding.
GenAI is neither intrinsically good nor terrible. The future is not black or white. Leaders should stay grounded, analytical, and deliberate in how they embrace GenAI and neglect extremes.
Chart of the week:
Battery storage in American homes jumped 64% from the year prior. Due to climate change-induced demand peaks and power outages, battery storage system installation is rising. Furthermore, solar panels and reducing battery prices are driving the trend.
These storage devices, which are currently installed in around 500,000 houses, have a total capacity of 3,328 megawatt-hours. Originally concentrated in sunny places like California and Texas, house battery installations are growing to other regions, including the Northeast and Mid-Atlantic.
Sound bites you should know:
Microsoft halts or delays data center projects in several locations. Is this the beginning of a reevaluation of AI and cloud infrastructure expenditure spreading to other hyperscalers?
Elon Musk announced the acquisition of X. Is Musk positioning xAI to compete head-to-head with OpenAI, Google DeepMind, and Anthropic?
A clinical trial showed Dartmouth's AI chatbot Therabot reduced depression symptoms by 51%, anxiety by 31%, and eating disorders by 19%, comparable to traditional therapy. What other problems will AI solve at scale?