- Horizon AI
- Posts
- AI to Give Banking a $340 Billion Boost 💹
AI to Give Banking a $340 Billion Boost 💹
+ DeepMind teaches AIs to Learn like Humans
Gina Acosta
December 05, 2023
Welcome to another edition of Horizon AI,
A recent McKinsey report suggests that integrating generative AI could significantly boost profits in the financial sector. Simultaneously, DeepMind continues to innovate, introducing an approach that shifts away from relying on extensive data. Instead, their AI learns tasks directly by observing humans.
Let's jump right into it!
Read Time: 3.5 min
Here's what's new today in the Horizon AI
A Chart to Kickstart Your Day 📈: Nvidia’s largest customers this year
AI to Give Banking a $340 Billion Boost 💹
‘Animate Anyone’ Turns Images into Incredibly Realistic Videos
AI Research: DeepMind teaches AIs to Learn like Humans
Trending AI Tool of the day: Create documents in minutes with Watto
AI Image of The Day🎨: Cyborgs relaxing in a meadow
The Latest in AI and Tech 💡
A chart to kickstart your day
Nvidia’s largest customers this year
Source: Omdia Research
Nvidia's H100s are the preferred choice for powering generative AI among tech companies. The disparity in shipment volume between Meta, Microsoft, and other players is quite significant.
AI News
MCKINSEY
AI to Give Banking a $340 Billion Boost 💹
Source: McKinsey
A new report from McKinsey suggests that generative AI could significantly boost profits across the banking sector. Utilizing these cutting-edge tools for increased productivity gains could add $340 billion annually to banks' earnings, equating to a 9-15% increase in operating profits.
Key details from the McKinsey analysis:
Corporate and retail banks stand to gain the most
Main profit drivers: AI automation of repetitive human tasks in areas like sales, marketing, software engineering
Up to 70% of business activities likely to have automated components in the future
Workforce impacts remain uncertain - jobs are unlikely to disappear rapidly but a portion will have automated elements
The conclusions are unambiguous - generative AI is set to provide a dramatic boost to productivity and profitability across banking. But effectively leveraging these tools requires investment, planning, and reskilling to adapt workforces to the AI-enabled future.
ALIBABA
‘Animate Anyone’ Turns Images into Incredibly Realistic Videos
Source: Alibaba Group
Researchers at Alibaba Group’s Institute for Intelligent Computing have developed a new generative video technique called 'Animate Anyone' that transforms still images into realistic videos.
Details:
This system builds upon prior image-to-video models such as DisCo and DreamPose but with substantially improved quality and detail retention capabilities.
It analyzes a source image, extracts elements such as facial features and clothing patterns, then generates a sequence of images mapping those details onto slightly varied poses
Some issues remain with parts like eyes and hands as well as side and rear views, but overall this represents major progress from previous attempts prone to artifacts and errors
The researchers indicate they are preparing the demonstration and code for future public release so people can evaluate Animate Anyone firsthand
While not the first of its kind, this new technique is one of the closest to achieving a plausibly realistic AI-generated video of the image provided
AI RESEARCH
DeepMind teaches AIs to Learn like Humans
DeepMind has taken a significant step in AI learning. Their new system of AI agents can learn tasks by directly observing humans, much like how you learn by watching video tutorials.
Details:
DeepMind's innovative approach, termed ‘Cultural Transmission’, eradicates the need for extensive human data. The AI system learns efficiently by observing and imitating human actions, resembling a few-shot imitation process.
By using real-time learning, the method could eliminate privacy concerns linked to storing large data sets for model training.
This novel learning method holds promise for robotics, streamlining tasks like lifting and placement. Additionally, it could enhance customer service systems with adaptive learning for personalized support.
Limitations: The agent may make incorrect generalizations about what it observed. Training scenarios were relatively simple; more diverse scenarios are needed to test if the technique would work more widely.
This latest research from DeepMind into imitative learning is a significant step towards training AI for real-world tasks more efficiently and privately.
Trending AI Tool
Create documents in minutes with Watto
Source: Watto
Watto can revolutionize your document creation experience, empowering you to 10x your productivity effortlessly.
Features:
Unified Knowledge Hub: Connect seamlessly across 10+ platforms, including Google Suite and Jira.
AI-Powered Document Creation: Generate expert-level documents quickly, freeing up your time for more strategic tasks.
AI Co-Pilot for Effortless Writing: Experience smooth writing with AI co-pilot features like rephrase and shorten.
20+ Templates: Start projects instantly with a diverse range of templates – from PRDs to One Pagers, ensuring professional and impactful documents every time.
Check this AI tool here
AI Image of The Day
Cyborgs relaxing in a meadow
Send your AI-generated images for a chance to get published here: [email protected]
Source: @PromptlyAI_YT on X
The latest in AI and Tech
Don't believe the AI hype, says OpenAI Exec
OpenAI COO Brad Lightcap cautions that while AI offers promise, it can't instantly solve business problems or deliver major change overnight. He notes AI is still experimental, and not yet reliable for critical uses. We're still discovering its potential. Rather than a magic bullet, he suggests regarding AI as an evolving set of tools to selectively apply where they may augment human intelligence.
OpenAI rival Mistral nears $2 Billion Valuation
Mistral AI, founded in 2023 by ex-researchers from Meta Platforms and Alphabet, is poised to secure a staggering $487 million in funding, bringing its valuation close to $2 billion. This Paris-based startup focuses on open-source software for generative AI, positioning itself as a cost-effective alternative to established players. Backed by investors including Andreessen Horowitz, Nvidia, and Salesforce, Mistral aims to challenge the dominance of US industry giants.
AI Alliance formed by Meta and IBM: A collective push for responsible innovation
Source: The AI Alliance
Industry giants Meta and IBM have co-launched the AI Alliance, a collaborative effort involving around 40 organizations from industry, government, and academia. Notable members include chipmaking leaders AMD and Intel, prestigious universities like Cornell and Yale, and key players in open-source AI tools such as Hugging Face, LangChain, and MLCommons. The alliance's objective is to cultivate an open community that promotes responsible AI research, pooling resources to address safety concerns and establishing benchmarks for AI system safety.
That’s a wrap!