TLDR AI 2024-07-16

In a recent interview, Microsoft CTO Kevin Scott expressed confidence that large language model "scaling laws" will continue to drive AI progress, countering skepticism about diminishing returns.

OpenAI recently shared a five-level system for tracking its artificial general intelligence progress. The levels go from what is currently available to AI that can perform the same amount of work as an organization. They may change based on feedback from employees, investors, and the company's board. OpenAI executives believe the company's AI is currently on the first level, but it is getting close to level two, where AI can perform basic problem-solving.

VC firm Andreessen Horowitz has secured thousands of AI chips to exchange for equity.

A great paper that outlines how you can turn a spreadsheet into a representation that is useful to a modern LLM. This can be used for Q/A, formatting, and other data operations.

Label-driven Automated Prompt Tuning (LAPT) is a new method for out-of-distribution (OOD) detection in Vision-Language Models like CLIP.

Hyper-3DG improves text-to-3D model generation by focusing on the complex relationships between geometry and texture.

ProDepth is a new framework for monocular depth estimation that tackles issues caused by moving objects in dynamic scenes. It uses a probabilistic approach to identify and correct inconsistencies in depth estimation.

LightenDiffusion improves low-light images using diffusion models and Retinex theory.

Open-Canopy is a high-resolution (1.5 m) open-access dataset for estimating canopy height across France.

Gem's 2024 Talent Acquisition Trends report dives straight into the chaos. Why are 55% of companies doubling down on AI and automation? Because they're tired of wasting time on mundane tasks. How are 68% of recruiters boosting their efficiency? By using smarter tools, not working harder. Want to know more. Read the full report here to get ahead and stay ahead.

An example of the high valuations being assigned to AI startups, Hebbia has raised $130 million at a $700 million valuation on $13 million ARR. Hebbia's main product, Matrix, can analyze large documents and respond to user inquiries in a spreadsheet-like format. It is currently used by 30% of asset managers for research purposes. The company plans to use the fresh funding to expand its team and offerings to other industries, such as law firms and pharmaceutical companies.

The debate over AI potentially achieving consciousness is intensifying, with contrasting views on whether advancements in machine learning and neuromorphic computing can lead to sentient machines. Integrated Information Theory suggests AI consciousness is unlikely due to current hardware limitations, whereas computational functionalist theories like Attention Schema Theory and Global Neuronal Workspace Theory consider it inevitable. Neuroscience seeks a unified theory of consciousness to better understand its potential emergence in AI.

There is an emerging paradigm of Managed-Service-as-Software (M-SaS) startups, where AI-driven service-oriented startups build their companies according to a new business model blueprint. This requires a fundamental shift in mindset for startups to use AI, rather than to sell AI. Initially labor-intensive with low gross margins, these startups gradually shift to higher SaaS-like gross margins through automation and AI.

A new toolkit built on many scraping softwares to help collect web data at scale.

The latest Google app beta reveals that buyers of the "Pixel Pro" series, likely including the Pixel 9 Pro and Pixel 9 Pro Fold, will receive a 1-year free subscription to Gemini Advanced, Google's most capable AI model.

HR software company Lattice faced backlash after announcing it would grant "rights" to AI algorithms, integrating them as "digital workers" with managers and performance assessments in its product.