Future of AI // AI of the future
There are no new ideas or entirely new science around artificial intelligence
It’s hard to imagine the progress that will be made in developing AI systems. //No, we are not talking about AGI or human-like intelligence, but the pace and scale of progress in practical AI systems are shockingly rapid. All we can do is mimic (or simulate) cognitive functions, but with modern interconnect, computing power, and memory, it’s possible to replicate any cognitive pattern and function. Modern systems are unable to synthesize new knowledge or solve new tasks yet, but researchers are working to overcome this.
Spatial intelligence // what the hack is this?
Here are a few key points I gathered from the conversation about spatial intelligence and World Labs:
1. Spatial intelligence refers to machines’ ability to perceive, reason about, and act in 3D space and time. It involves understanding how objects and events are positioned and interact in the physical world.
2. World Labs is focusing on spatial intelligence as a new frontier in AI, beyond language models and 2D image/video generation. The founders believe it’s fundamental to how intelligent beings interact with the world.
3. Key technical advances enabling this work include:
— Increased compute power
— Better understanding of data beyond just labeled images
— Algorithmic breakthroughs like Neural Radiance Fields (NeRF)
4. Potential applications include:
— Generating interactive 3D virtual worlds
— Augmented reality interfaces
— Robotics and physical world interaction
5. The team sees this as a deep tech platform that could enable many applications, rather than focusing on one specific product initially.
6. Challenges include the multidisciplinary nature of the problem, requiring expertise in AI, computer vision, graphics, engineering, etc.
7. Success for World Labs would mean their spatial intelligence models being widely used to unlock new capabilities across many industries and use cases.
The founders believe spatial intelligence represents a fundamental new capability for AI systems that could enable transformative new applications, particularly as AR/VR technologies mature. They are taking a long-term view on developing this technology as a platform.
World Labs vacancies: Graphics Engineer, Hardware Engineer, Product Engineer, Product Manager, Research Engineer, Research Scientist, Systems Engineer // there are no scientific roles
AI Google vision
Here’s a summary of the key points from the conversation with Sundar Pichai, CEO of Google and Alphabet:
1. AI as a transformative technology:
- Pichai sees AI as the most profound platform shift of our lifetimes, comparing it to previous shifts like the internet and mobile computing.
- He believes AI will impact every industry and domain, from climate to health to education.
2. Google’s AI initiatives:
- Google has been investing in AI early on, with breakthroughs like the Transformer architecture.
- The company is now focused on its Gemini family of models, pushing frontiers in multimodality, long context, and reasoning.
3. Responsible AI development:
- Pichai emphasized the importance of developing AI responsibly from the start.
- Google follows AI principles published in 2018 and focuses on technological solutions, partnerships, and expanding access to AI benefits.
4. Energy and sustainability:
- Pichai acknowledged the energy demands of AI but expressed optimism about medium to long-term solutions.
- Google is investing in renewable energy sources and working on more efficient AI models.
5. AI governance and regulation:
- Pichai supports a balanced approach to AI regulation, emphasizing the need for innovation while addressing potential risks.
- He suggested leveraging existing industry regulations while allowing for AI innovation.
6. Impact on jobs and education:
- Pichai believes AI will augment human capabilities rather than replace jobs outright.
- He stressed the importance of reskilling and lifelong learning to adapt to rapidly changing skill requirements.
7. Google’s presence in Pittsburgh:
- Google Pittsburgh has grown to over 700 employees, working on various projects including search, ads, cloud, and AI research.
- Pichai praised the local ecosystem and its connection to universities like Carnegie Mellon.
8. Advice for students:
- Pichai advised students not to stress too much about immediately figuring out their career path.
- He encouraged them to take time to discover what they truly enjoy doing, as passion often leads to success.
Throughout the conversation, Pichai expressed optimism about AI’s potential to benefit society while acknowledging the need for responsible development and addressing challenges like energy consumption and ethical concerns.
Here’s a summary of Stephen Fry’s keynote speech on artificial intelligence:
1. Introduction:
— Fry acknowledges his lack of expertise but long-standing interest in AI
— He warns against expecting definitive answers or predictions
2. Historical Context:
— Traces AI development from Alan Turing to present day
— Highlights the importance of increased computing power and big data
3. Technological Impact:
— Compares AI’s potential impact to other transformative technologies
— Discusses the concept of “disruption” in various industries
4. Human vs. Machine Intelligence:
— Explores the differences between human and artificial intelligence
— Questions what uniquely human qualities might remain valuable
5. Challenges and Risks:
— Addresses the “alignment problem” in AI development
— Warns of potential misuse by countries, corporations, and criminals
6. Regulation and Control:
— Compares AI regulation needs to those of money/financial systems
— Discusses the European Union’s AI Act as a potential regulatory model
7. Call to Action:
— Urges for immediate establishment of “red lines” in AI development
— Emphasizes the need for mandatory AI self-disclosure
8. Conclusion:
— Stresses the urgency of channeling and controlling AI’s development
— Echoes the Russell-Einstein Manifesto, calling on humanity to act responsibly
Throughout, Fry uses metaphors of rivers, tsunamis, and rising tides to illustrate the gradual but potentially overwhelming impact of AI on society. He emphasizes the difficulty in predicting AI’s future effects and the importance of proactive, global cooperation in managing its development and implementation.
AI and scientific discovery
SciAgents is an AI system designed to drive scientific discovery by integrating large-scale knowledge graphs, large language models (LLMs), and adversarial interactions between expert agents. It autonomously explores novel domains, identifies complex patterns, and uncovers hidden connections in vast scientific data. Key features include:
1. Ontological Knowledge Graphs: These structure and connect scientific concepts across disciplines.
2. Multi-Agent Collaboration: AI agents generate, critique, and refine hypotheses while evaluating trends.
3. Graph-Based Reasoning: This enables the discovery of novel materials by identifying interdisciplinary connections, such as biomimetic designs inspired by natural patterns.
SciAgents can function autonomously or collaboratively with human researchers, offering powerful data processing and innovative paths for exploring nature-inspired designs. In materials science, it has already demonstrated how principles from various fields, including biology and music, can converge to create new materials. The system enhances researchers’ capabilities, allowing them to explore large datasets and develop hypotheses grounded in a vast, interconnected knowledge web.
Chat-GPT-o1 is reported to have achieved step 2 of 5 in OpenAI’s roadmap towards AGI. It’s already scoring an estimated IQ of 120, surpassing the average human in some tests. Is AGI achievable by end of decade?
