AI Weekly Buzz: Innovations, Insights, and Industry Trends August 25th — 31st
Hey! Is it just me, or was August 2024 a fun ride? New models, updated AI tools, controversial AI applications and several thought-provoking occurrences graced the month. Welcome to another AI Weekly Buzz. Here is a look at major AI happenings in the last week of August 2024.
Google AI Doom Simulation
GameNGen (pronounced Game Engine) is a neural network (AI) powered environment that can simulate in high quality and real time, the game DOOM (a classic released in 1993) at over 20 fps on one TPU. The researchers from Google DeepMind, Google Research and Tel Aviv University in their study say
GameNGen is trained in two phases: (1) an RL-agent learns to play the game and the training sessions are recorded, and (2) a diffusion model is trained to produce the next frame, conditioned on the sequence of past frames and actions.
What does this mean for the Game industry? Better still, what progress can we expect in video generation and simulation in the coming days? This research proves that interactive software such as a game can be transformed into a neural model to generate something new that humans do not program. The researchers suggest that this could lead to cheaper and faster productions of new games and the potential for games that are dynamic to individual, with scenes or characters being created on the fly.
The study also shows that there are limitations involved. These include a limited amount of memory (a little over 3 seconds history/context length for the next video frame) and a difference between the behaviour of the agent and the human players.
GameNGen has caught the attention of both AI enthusiasts and gamers. Hopefully, this will lead to meaningful and better user experiences with video games and more advancements in life experiences with interactive software systems.
California’s AI Safety Bill
The Safe and Secure Innovation for Frontier Artificial Intelligence Models Act was recently amended and approved by the California State Assembly. It is still to be voted in the State Senate, but from observations, it is expected to pass. Some of the notable requirements of the bill include:
- Implement reasonable administrative, technical, and physical cybersecurity protections
- Implement the capability to promptly enact a total shutdown.
- Implement a written and separate safety and security protocol
- Conduct an annual review of the safety and security protocol to account for any changes
- The developer shall conspicuously publish a redacted copy of the auditor’s report and transmit to the Attorney General a copy of the redacted auditor’s report.
Baidu hinders Google and Bing
Baidu, a large and widely used search engine in China, blocked other online search engines like Google and Bing from indexing its content. This is to safeguard its assets as the demand for data to train and build models has increased.
This is similar to Reddit’s decision in July to exclude Google from its search engine list. Google made a multi-million deal with Reddit and is authorized to scrape the site. Another similar partnership is the relationship between OpenAI and Financial Times.
Baidu’s step exposes the increasing importance of data in the AI age. As various companies delve into AI and its applications, data-sharing policies will continue to change. We may be seeing more of this kind of Baidu’s restriction in coming days.
AI and Data Science in AIT Worldwide Logistics
“AIT is continually looking towards the future, With the establishment of the AI and data science team, the company is prepared to capitalize on generative AI, natural language processing, machine learning, large language models, and other advancing technologies.”
Vaughn Moore, Chairman and Chief Executive Officer
The AI and data science group consists of three primary teams:
- Data Enablement: This team spearheads a structured, company-wide approach to transform existing data silos into next-generation-compatible assets, fostering new levels of trust with stakeholders by enhancing efficiency and continually improving solutions.
- AI Governance and Data Science: Focused on creating innovative AI business features, this team prioritizes increased accuracy and robust privacy protections for customers and colleagues as technology advances and evolves.
- Business Intelligence and Analytics: This team is responsible for delivering comprehensive, centralized reporting and visualization services across the organization. This team provides thorough, centralized reporting and visualization services across the organization.
Aviation — AI in Pilot Training
With all the advancements birthed by artificial intelligence in various sectors, Embry‑Riddle Aeronautical University decided to utilize AI to enhance the students’ training.
Here are some of the ways AI is applied
- AI’s automatic speech recognition, speech-to-text conversion, and natural language processing enable it to interpret and respond to human speech, providing real-time corrective feedback.
- Advanced Simulation Technology Inc. (ASTi) has developed systems specifically designed to replicate the aircraft, airspace, and flight scenarios that students will experience.
- Combining a customized VR flight trainer and ASTi’s Simulated Environment for Realistic ATC (SERA) utilizes machine learning to create a dynamic environment featuring simulated air traffic controllers and radio communications from other aircraft.
- The University of Iowa’s Operator Performance Laboratory (OPL) offers an AI tool that is trained on data from both novice and expert pilots, capable of detecting levels of flight skill mastery.
“This tool can create scenarios and lessons that reinforce training in areas where the student is weaker”
said Dr. Tom Schnell, Director of OPL
Education sector — UK Funds AI in Schools
The UK government recently invested £4 million in the AI initiative towards improving the education sector. This move will target curriculum, assessments and other facets of schools, using AI tools to improve the system.
We know teachers work tirelessly to go above and beyond for their students.
By making AI work for them, this project aims to ease admin burdens and help them deliver creative and inspiring lessons every day, while reducing time pressures they face.
This is the first of many projects that will transform how we see and use public sector data. We will put the information we hold to work, using it in a safe and responsible way to reduce waiting lists, cut backlogs and improve outcomes for citizens across the country.
- Peter Kyle, Science Secretary
We are determined to break down the barriers to opportunity to ensure every child can get the best possible education — and that includes access to the best tech innovations for all. Artificial Intelligence, when made safe and reliable, represents an exciting opportunity to give our schools leaders and teachers a helping hand with classroom life.
Today’s world-leading announcement marks a huge step forward for AI in the classroom. This investment will allow us to safely harness the power of tech to make it work for our hard-working teachers, easing the pressures and workload burdens we know are facing the profession and freeing up time, allowing them to focus on face-to-face teaching.
- Stephen Morgan, Minister for Early Education
Word on the Streets
OpenAI Strawberry Season is Coming
OpenAI has been close-mouthed about this, but according to Reuters, OpenAI is working on a new reasoning project code-named “Strawberry”, formerly known as Q*. Some of the expectations from this project include:
- “Deep research " means that the model will be able to browse the Internet autonomously, making better findings and fewer hallucinations. This will be done with the assistance of a computer-using agent (CUA) to go beyond findings and take action on these research findings.
- Possibly better mathematical capabilities with over 90% on the MATH dataset benchmark
- Super-human-level reasoning with better abilities to detect logical fallacies and solve common sense problems.
A spokesperson of OpenAI was asked about Strawberry, and instead of addressing it directly, he said
“We want our AI models to see and understand the world more like we do. Continuous research into new AI capabilities is a common practice in the industry, with a shared belief that these systems will improve in reasoning over time.”
…
That’s it for this week. Hopefully, you found the news intriguing. Watch for next week’s edition of AI Weekly Buzz, where more innovations, insights and trends will be delivered. If you have not, kindly check out our other articles, too.
À bientôt!
