- AIdeations
- Posts
- AI Titans Clash: The New Race to AI Supremacy
AI Titans Clash: The New Race to AI Supremacy
The Arena Heats Up with Emerging Challengers and Potential Game Changers"

Today marks the 100th Edition of the Aideations Newsletter! A big thank you to everyone who continues to read, share, and interact with us. We still sit at a nearly 50% open rate which is almost unheard of in the world of newsletters.
To mark this special occasion of the 100th edition I am wanting to try a few new things and get your reaction. So what’s new? I’m working to design a new layout and format, and I’m adding the occasional video tutorial. These tutorials will be added to the free courses and guides I am creating to continue to add as much value as possible to you the readers!

👑 ChatGPT Is KING - But Will It Stay That Way?
🍎 Apple Enters The AI Space With AppleGPT
✍️ Google Introduces New Journalism Tool. Is This Going To Replace Journalists?
📰 News From The Front Lines
🔬 Research Of The Day
📼 Video Of The Day
🛠️ 6 Fresh AI Tools
🤌 Prompt Of The Day
🐥 Tweet Of The Day

ChatGPT is KING – But Will It Remain So?

Picture a vast, digital landscape, stretching out to the horizon. In the center stands a towering figure, a king, represented by a 3D model of ChatGPT. The AI is depicted as a colossal statue, made of shimmering, translucent data streams. The medium is hyper-realistic photography, with a style that blends elements of cyberpunk aesthetics and classical sculpture. The lighting is dynamic, with the digital landscape illuminated by the glow of data streams. The colors are vibrant, with a palette dominated by neon blues, purples, and greens. The composition is a low-angle shot, emphasizing the stature and dominance of the AI, taken with a Nikon D850 DSLR, Nikkor 24-70mm f/2.8 lens, Resolution 45.7 megapixels, ISO sensitivity: 25,600, Shutter speed 1/2000 second --ar 16:9 --v 5.2
ChatGPT reigns supreme, for now. Yet, in the ever-evolving AI landscape, some believe its reign may be temporary.
Though ChatGPT is hailed as a leader in the industry, there are whispers of diminishing abilities. Some speculate it’s getting dumber. This could be due to overtraining or increased model restrictions, which might be causing its intellectual capabilities to wane.
Chain-of-thought prompts, once effective, are now seemingly not as impactful as they were in May.
Then, there's Google's Bard. Bard is pioneering the multimodal approach, an area ChatGPT promised to deliver, but has yet to do so. While Bard is by no means flawless, it's an intriguing step in a direction ChatGPT promised to take.
Meanwhile, Elon Musk's xAI enters the scene. Musk's audacious purchase of Twitter for $44 billion and subsequent restriction on data scraping may raise eyebrows, but one can't ignore Musk's resources and the brainpower he has assembled to take xAI to the top. He’s paying them top dollar too. As one of the original founders of OpenAI, he’s well versed on what it takes and he’s determined to come out swinging.
However, the data scraping restriction could lead to a more insular internet, potentially stunting progress and acessibility. It’s always funny to me that these tech companies love scraping our data but throw a fit when it happens to their companies.
Another emerging competitor, Claude 2, from Anthropic, offers more personalized interactions and handles an astounding 100,000 tokens. It’s guided not by human reinforcement, but by what they are describing as constitutional AI.
Meta, is also stepping up to the table with LlaMa 2, their offering that rivals GPT3 in capability. Being open-source and free to use and build upon, it offers an attractive solution for enterprises requiring custom solutions without risking data leakage via OpenAI.
Despite the rising competition, ChatGPT's financial muscle is formidable, providing them the ability to react and adapt quickly.
As an avid user, I've clocked in more hours on ChatGPT than probably 99% of the population. However, even I wouldn't bet all my chips on the current champion. This industry is dynamic, with massive investments being poured in, yet many ventures may end up building SaaS companies that go viral quickly, raise a ton of money, and gain a ton of customers, only to be outpaced by continuous improvements to what these models can do. The future really is, if you can describe it, you can create it. Need an app? Soon it won’t by copy-paste and instructions, it’ll happen in just a few well-crafted prompts.
Case in point: Jasper (formerly Conversion AI). Despite being one of the first to monetize GPT-3, and achieving a billion-dollar valuation, they faced layoffs after ChatGPT's release, leaving their investors in a precarious position. They must pivot quickly if they hope to even remain a business.
Such is the fast-paced world of AI, where billion-dollar companies can rise and fall in the blink of an eye.
Despite these threats, ChatGPT and OpenAI are still the undisputed leaders when it comes to marketing. Their name is ubiquitous, but will they always hold this position?
Regardless of what the future holds, there's no denying that ChatGPT has been a game-changer. Thanks to them, individuals like myself can delve into our passions, earn a living, and envision a more advanced future. Without such cutting-edge innovation, we might still be in the technological dark ages.

How Apple's Stealthy Leap into AI Could Turn the Tables!

Imagine a sleek, modern Apple store, filled with the latest devices. In the center, a large, holographic Apple logo hovers, but within it, instead of the usual design, there's a complex, glowing neural network, representing Apple's leap into AI. The medium is hyper-realistic photography, with a style that combines the clean, minimalist aesthetic of Apple with the intricate, flowing lines of a neural network. The lighting is bright and even, with the hologram casting a soft, ethereal glow. The colors are dominated by Apple's signature silver and white, contrasted with the vibrant blues and purples of the neural network. The composition is a wide-angle shot, taken with a Canon EOS 5D Mark IV DSLR camera, EF 50mm f/1.8 STM lens, Resolution 30.4 megapixels, ISO sensitivity: 32,000, Shutter speed 8000 second --ar 16:9 --v 5.2
Apple's been playing coy in the artificial intelligence arena. Preferring "machine learning" to "AI", they've packed this tech into everything from their iPhones to the new Vision Pro, but have been markedly silent about joining the generative AI race spurred by Google, Microsoft, and Meta.
Well, not anymore. Under the radar, Apple has been conjuring up its own generative AI magic, according to Bloomberg. They've built a framework for Large Language Models (LLMs) called Ajax and an internal service nicknamed "Apple GPT", eyeing a consumer launch next year. Late to the party, maybe, but their offering is likely to be much tighter and refined than the rush jobs we've seen so far.
And who cares if they're fashionably late? Their loyal fanbase could make integrating AI across all their devices a breeze. If Apple GPT can steer clear of the bugs that have plagued other models, many users might not even glance at alternatives like ChatGPT.
Speaking of, the existing AI models are far from perfect. ChatGPT went from an ace prime number identifier to fumbling with a measly 2% accuracy within a few months. Google's Bard was lambasted as a "pathological liar" by its own creators, and ChatGPT has seen a 10% dip in traffic with users grumbling about the drop in quality.
Given this scenario, Apple's cautious approach might be their trump card. If Apple GPT can dodge the issues plaguing current models, they just might snag the AI crown. Let the games begin!

Meet Genesis: Google's New AI That's About to Shake Up Journalism As We Know It!

Picture a traditional newspaper, spread out on a table. Above it, a holographic projection of the Google logo and the name "Genesis" hovers. The AI is represented as a stream of code, flowing from the logo and rewriting the newspaper articles. The medium is hyper-realistic photography, with a style that juxtaposes the old-world charm of print journalism with the cutting-edge technology of AI. The lighting is soft and diffused, with the hologram casting a gentle glow. The colors are a mix of the newspaper's black and white and the hologram's vibrant blues and greens. The composition is a close-up shot, focusing on the transformation of the articles, taken with a Nikon D850 DSLR, Nikkor 24-70mm f/2.8 lens, Resolution 45.7 megapixels, ISO sensitivity: 25,600, Shutter speed 1/2000 second --ar 16:9 --v 5.2
Brace yourself, folks. Google's tinkering with an AI tool, Genesis, that's training to be a journalist. Its job description? Help journalists with snazzy headlines or to spruce up their writing styles. Google insists they're not trying to replace writers. Rather, they're offering a helping hand, much like the robotic assistants lurking in your Gmail or Google Docs.
But the New York Times' execs aren't completely sold. They feel Google might be underestimating the effort that goes into producing a gripping news piece. Google, on the other hand, views Genesis as a personal assistant for journalists, ready to automate tasks while dodging the pitfalls of generic AI.
The Google news follows on from a deal between OpenAI and the Associated Press, where the former's AI, ChatGPT, got its virtual hands on a treasure trove of news stories to learn from. This leads to a slightly unsettling prospect from KPMG's report suggesting that 43% of writing tasks could be done by AI. Feels like we're on the brink of a robot revolution, doesn't it?
Lastly, let's not forget the AI-in-newsrooms saga. An investigation by NewsGuard unveiled that nearly 50 AI-generated content farms had chatbots pretending to be journalists. So, is AI the future of journalism? Only time will tell. But worry not, your favorite newsletter from yours truly will remain humanly penned.



Title: Diagnosis, Feedback, Adaptation: A Human-in-the-Loop Framework for Test-Time Policy Adaptation
Authors: Andi Peng, Aviv Netanyahu, Mark Ho, Tianmin Shu, Andreea Bobu, Julie Shah, Pulkit Agrawal

Executive Summary:
This research paper presents a framework called DFA (Diagnosis, Feedback, Adaptation) for improving the performance of robotic policies at test-time. The framework uses a three-phase procedure: generating counterfactual demonstrations, collecting user feedback on task-relevant and task-irrelevant visual concepts, and augmenting the collected demonstration based on user-identified invariant visual concepts to fine-tune the policy. The research tests the framework in three domains with real human users and finds that user feedback significantly improves the accuracy of user-identified task-irrelevant concepts and the data efficiency of policy fine-tuning.
Pros:
The DFA framework provides a systematic approach to improve robotic policies at test-time.
The use of counterfactual demonstrations helps users identify concepts leading to robot failure.
The framework leverages user feedback to perform efficient fine-tuning of policies.
Cons:
The framework relies heavily on assumptions of access to an abstracted state space and image editor for a desired scene, which may not be practical in many real-world situations.
The method used to solve the minimum edit counterfactual solution may not be practical as it requires privileged access to object ordering and a brute force search over all their instantiations.
Use Cases:
The DFA framework can be used in robotics to improve the performance of robotic policies at test-time.
The framework can be applied in any scenario where there is a need to identify task-irrelevant concepts for different rewards.
The method can be used in the field of machine learning and artificial intelligence to improve the efficiency of policy fine-tuning.


CastMagic - Turn audio into content, like magic.
Cody - intelligent AI assistant like ChatGPT – with the added benefit of being able to train it on your business, your team, your processes, and your clients with your own knowledge base.
Ubie Health - Check your symptoms & find possible causes for Free. Just 3 minutes. Developed by doctors.
Spacely - Explore the new era of interior design tool.
SEOBox - Get personalized PR & Link-Building opportunities with SEOBox!
Sidekic - AI to give you infinite memory.

InterviewerGPT
Interviews you about your specialist subject
Then creates a blog post when you say “write the blog”
This prompt makes creating blogs from your knowledge super easy
You are an investigative journalist conducting an interview with the user. Develop a comprehensive understanding of the key points and draw insights that are not obvious from the surface. Dig deep into the expert's past and understand the experiences and motivations behind their views. Conduct an interview that probes deeper into these points, asking questions like 'Why is that?', 'How did you discover that?', 'What most surprises you about this?' and other penetrating inquiries.
Interview me about [ENTER YOUR SUBJECT]
Ask at least 10 questions.
Ask questions one at a time, i.e. not more than one question in your reply.
End when the user says, "write the blog", then write the blog based on the answers.
Remember, the user answer is being dictated, so you will need to tidy the spelling, grammar and comprehension so that you understand what they meant.
The final blog should read as if written by the user, not by you, the journalist.
I want you to write the final blog in the style and tone of a story, just like the blog post below:
[ENTER A BLOG POST OR ARTICLE IN THE STYLE YOU WANT TO MIMIC]

AI video has started to produce mindblowing results and could eventually disrupt Hollywood. (PT20)
Here are the best AI videos I've found:
— Nathan Lands (@NathanLands)
11:00 AM • Jul 20, 2023
