Research & Development World

This week in AI research: OpenAI launches full ‘reasoning’ model and $200-per-month plan

By |

This week in AI brings both breakthroughs and reality checks. DeepMind’s weather forecasting model demonstrates high accuracy at 15-day predictions, potentially transforming climate science and disaster preparedness if it can live up to the billing. Meanwhile, OpenAI’s new o1 model launch reveals the current limitations of AI reasoning — despite performance gains on academic tasks, the system stumbles on practical applications like woodworking instructions. In policy developments, Trump’s appointment of David Sacks as AI czar signals a potential shift toward lighter-touch regulation of AI and crypto. The media landscape sees its own AI controversies, with the LA Times planning an AI “bias meter” that has prompted staff resignations and debates over algorithmic oversight of journalism. In enterprise AI, Microsoft’s research team previews significant changes coming in 2025, including the emergence of AI agents that can handle complex organizational tasks like employee networking. As I wrote about here though, adoption is likely to be gradual and phased.

Featured stories


Click on the headlines to see the full preview.


Policy & Regulation
→

Former PayPal COO David Sacks named White House AI and crypto czar by Trump.


Media & AI
→

LA Times owner plans AI “bias meter,” leading to newsroom controversy.


Research & Innovation
→

OpenAI launches ChatGPT Pro at $200/month, offering advanced reasoning capabilities and improved reliability.


Enterprise & Market Trends
→

Amazon announces a massive AI supercomputer and new custom AI server, challenging Nvidia’s GPU dominance.


The 4th Wave of AI Drug Discovery
→

The “4th Wave” of AI-driven drug discovery emerges, reshaping pharma R&D.


LinkedIn Insights & Community News
→

OpenAI’s new office in Zurich signals growing European AI presence, plus other community highlights.


More Stories
→

Additional coverage including SoundHound AI stock surge, UHC AI denials, and academia’s AI concerns.

Policy & regulation

Sacks

David Sacks [Craft Ventures headshot]

Source: Reuters
Trump appoints former PayPal exec David Sacks as AI and crypto czar (December 6, 2024)

Summary:President-elect Trump named former PayPal COO David Sacks as “White House A.I. & Crypto Czar.” The role aims to create a regulatory framework for crypto industry clarity. With support from tech leaders including OpenAI’s Sam Altman, Sacks is expected to balance necessary guardrails with minimal red tape to foster innovation. The price of Bitcoin recently surpassed $100,000. Sacks has invested in both cryptocurrency and AI.

Media & AI

Dr. Patrick Soon

Dr. Patrick Soon-Shiong [NHS Confederation via Wikimedia]

Source: CNNLA Times owner plans AI-powered ‘bias meter’ on news stories (December 5, 2024)

Summary: LA Times owner Patrick Soon-Shiong announced plans to implement an AI-driven “bias meter” on news articles, offering readers “both sides” of coverage. The January rollout has sparked internal controversy and resignations among key newsroom figures who question algorithmic neutrality in editorial decision-making. Soon-Shiong acquired the LA Times in 2019. Soon-Shiong has criticized the blurred lines between news and opinion coverage generally.

Research & innovation

Source:
OpenAI
Introducing ChatGPT Pro (December 5, 2024)

Summary: OpenAI’s ChatGPT Pro subscription plan ($200/month) provides access to the most advanced OpenAI models, including “o1 pro mode,” which offers improved reliability and longer reasoning. While performance is higher on benchmarks, some logical errors remain in real-world tasks.
OpenAI

[OpenAI]

Source:
Nature
DeepMind AI weather forecaster beats world-class system (December 4, 2024)

Summary: Google DeepMind’s AI model outperforms top-tier forecasting systems, producing accurate 15-day predictions.
DeepMind weather forecasting

[Image from DeepMind announcement]

Source:
TIME
OpenAI’s New Ad Shows ‘Reasoning’ AI Making Basic Errors (December 6, 2024)

Summary: Despite improvements, OpenAI’s demo highlights that even advanced AIs struggle with practical reasoning, reflecting the technology’s current limitations.Birdhouse instructions

Enterprise & market trends

Source:
Wall Street Journal
Amazon Announces Supercomputer, New Server Powered by Homegrown AI Chips (December 3, 2024)Amazon’s massive AI “Ultracluster” and new “Ultraserver” with Trainium chips aim to cut training costs and challenge NVIDIA’s market dominance.

Source:
Yahoo Finance
Big changes are coming to AI in 2025: Microsoft (December 6, 2024)Summary: Microsoft anticipates advanced AI capabilities by 2025, improving reasoning, enterprise functionality, and employee networking tools.
Source:
Amazon Press Release
Amazon Bedrock Empowers Customers to Accelerate Generative AI Adoption (December 4, 2024)
Summary: Amazon’s Bedrock platform adds over 100 models for simpler enterprise generative AI deployments.
Meta LogoSource:
Quartz
Meta is joining Big Tech’s AI rush to nuclear power (December 6, 2024)
Summary: Meta seeks nuclear energy partners for future data center demands, reflecting tech giants’ move to secure stable, low-carbon power sources for energy-hungry AI operations.

LinkedIn community news

Source:
LinkedIn

  1. OpenAI Expands in Europe: New office in Zurich. More info
  2. LLM Hackathon Preprint: View Preprint
  3. Drug Discovery’s 4th Wave: Mentioned by Andrii Buvailo, Ph.D. (see below for more info)
  4. ML vs. AI vs. GenAI Experts: View Graphic
  5. TechBio News: Newsletter

The 4th Wave of AI Drug Discovery

Source:
LinkedIn Post by Andrii Buvailo, Ph.D.
The “4th Wave” of AI Drug Discovery is Here (December 3, 2024)

Summary: A new report highlights a shift toward holistic, biology-first approaches in AI-driven drug discovery, promising more transformative breakthroughs across pharma R&D. Buvailo has the scoop on the report, which sees the field of AI-driven drug discovery is entering its fourth evolutionary wave. The wave involves a fundamental shift from narrow, target-based approaches to holistic, biology-first methodologies. This new paradigm emphasizes the development of comprehensive “world models” of cellular and multicellular systems, tapping advanced AI technologies such as unsupervised learning and foundation models, combined with sophisticated laboratory automation and extensive biobanks.

Research assistance: Frédéric Célerse, Ph.D., Andrii Buvailo, Ph.D.

Related Articles Read More >

Search R&D World