Edward Frank Morris

📣 New Research: The State of AI Report 2024, produced by AI investor Nathan Benaich and Air Street Capital, analyses the most interesting developments in AI. This report is the biggest and most comprehensive yet, covering everything you *need* to know about research, investment, industry, safety and politics. Highlights this year include: ▪ The state of play on foundation models, including the performance of the big labs, smaller models and attempts to manage data and resource constraints ▪ The latest in hardware, the chip wars, and Chinese labs’ ability to defy US constraints - Fundraising numbers for VC across different AI categories and geographies, as AI companies reach $9T in enterprise value ▪ The sectors where sustainable AI businesses are emerging and gaining traction, along where progress is slower ▪ The politics of AI, including regulatory developments, copyright, deepfakes, infrastructure, and energy use ▪ The cooling of the existential risk debate and the development of technical safety literature. Download the complete report here (It's free, no registration needed): https://www.stateof.ai/ #stateofaireport #ArtificialIntelligence #GenerativeAI #genAI #EnterpriseAI #AIadoption #AIdevelopment #StateofAI #2024report #researchreport #technologydeployment #innovation #airace #airally #businessbenefits #aifuture #aiimpact #businessimpact #aipowered #aienabled #aiguided #aibusiness #aistartups #aistrategy #technologyadoption #technologyuse #opensourceai #disruptivetech #AImodels #multimodalAI #aitrends #technologyadvancements #artificialintelligencetechnology #ai4good #aiapplications #technologyinnovation #llm #ml #aievolution #aidriven #humantouch #aiadvancement #airesearch #business #companies #industries #aichallenges #businessgrowth #digitaltransformtion #aitransformation #llms #ml #AIOps #2025outlook #insights #politics #china #aigovernance #airegulation #aicompetition #aichips #aihardware #AIfirst #competitivelandscape #openai #agenticAI #opensource #businesscase #multimodalai #futureofai #airevolution #aivalue #aiethics #aisafety #agi #aithreats #airisks #vc #investors #AIinvestments #venture #pe #funding #fundraising #capitalraising #freedownload #freereport #pseudoacquisitions #mergers #acquisitions #businessvaluation

12

3 Comments

The United States is ahead of China, the EU, and the U.K. as the leading source of top AI models. In 2023, 61 notable AI models originated from U.S.-based institutions, far outpacing the European Union’s 21 and China’s 15. According to Crunchbase, venture funding to AI-related startups actually increased in Q1 2024 compared to Q4 2023. Around $12.2 billion were invested in venture-backed AI startups in 1,166 deals in the first quarter of 2024. Sources in the first comment. 

18

3 Comments

🚨 Microsoft and Google are fighting over the future of Xbox Microsoft and Google are currently in a legal dispute that could impact the future of the Xbox app on Android. Microsoft has developed a new Xbox game store for Android but is unable to launch it due to a court ruling that affects Google’s Play Store policies, specifically regarding app billing requirements. Microsoft is seeking regulatory changes to allow direct sales and streaming of games through its app without incurring Google’s 30% fee on purchases. Google has countered Microsoft’s claims, stating that Microsoft has the ability to sell games directly through its app but has opted not to do so. Historically, Microsoft removed the purchasing feature from its Xbox mobile app in 2020 to focus on remote console streaming. The legal battle is crucial for Microsoft's strategy to extend Xbox beyond traditional consoles, as it aims to create a more integrated gaming experience across devices. The outcome of this dispute could significantly shape the mobile gaming landscape and Microsoft's ambitions in that market. 💬 Source

1

The AI Loop: Machines Optimizing Content for Machines In today's InnovateUK BridgeAI session, Guy Gadney, FRSA highlighted a fascinating trend: ChatGPT is being used to generate content specifically optimized for YouTube's algorithms. This content is then further analysed and refined by YouTube's machine learning marketing models. Essentially, we have machines creating content that's perfectly tailored to fit machine-designed algorithms. It's a full circle of AI optimization. This raises intriguing questions: 1. What does this mean for the future of content creation? 2. Will human creativity still play a role? 3. How will this impact the way we consume and interact with content? What are your thoughts on this AI loop? Share your insights below! #AI #ContentCreation #YouTube #Algorithms #MachineLearning #FutureofWork #InnovateUK Disclaimer: partially written with a prompt using Gemini for amusement- my LinkedIn content is written by a human unless specified otherwise

3

Is this AI? Here’s how you can tell if someone is using LLMs👇 This is a list of “gpt tells” that I’ve been keeping: 1. American English (in the UK) Specialize v specialise, optimize v optimise 2. Being too profound: In the ever-evolving landscape / in the digital age / in the dynamic world of etc. 3. Adding conclusions when there is no need for one. 4. Emails starting with “I hope this finds you well”. 5. CVs with: Streamlined / Unwavering / Hone my ability 6. Rookie errors: Absolutely! / Certainly! 7. Other stuff: Complex and nuanced Multifaceted Delve Tapestry Multitude Unwavering Embark on a journey Testament Unleash as we journey through navigating through as we delve deeper Moreover / Furthermore Myriad Kaleidoscope Have I missed any? I’ve noticed certain words and phrases popup for certain fields. So it would be interesting to see yours too. #chatgpt #intheeverchanginglandscape

8

12 Comments

30% of people in the UK have not heard of any of the most popular AI tools (including ChatGPT) despite nearly two years of hype. This is according to a research by the Reuters Institute and University of Oxford. ---- ↳ The public has nuanced expectations when it comes to the impact of generative AI across different sectors. ↳ While many expect large impacts and believe sectors like healthcare and science will use it responsibly, there is skepticism around players like - Social media companies - Politicians - And news media outlets ↳ Despite the hype and media coverage around generative AI, a sizable 19-30% of the online population across six countries surveyed were not familiar with any of the popular AI tools like ChatGPT. ↳ And while many have tried using them experimentally, frequent and routine usage remains quite rare so far. ↳ When it comes to journalism specifically, the public primarily expects generative AI to help publishers by reducing costs of production. ↳ However, they identify few ways it will actually improve the - Quality - Trustworthiness - Or transparency of news for audiences. ↳ Very few think AI-generated news will be worth paying for compared to journalism by humans. ↳ There is a perception among a substantial minority that journalists already rely heavily on generative AI for tasks like - Grammar checking - Writing articles and headlines - Creating images - And even generating artificial authors or presenters. ↳This prompts discomfort about lack of transparency. ↳ Across many questions, a large segment of respondents (often one-third to one-half) opted for neutral or "don't know" responses ---- P.S. What are your thoughts? Sign up to my newsletter: https://bit.ly/4aQFHxJ

28

35 Comments

Chat GPT 4 Canvas - an amazing new Beta 😎 Below is a video showing how it works. The new interface allows you you to edit on the fly, without having to go back and edit the original prompt. It is great for content creation but will also analyse blocks of code and make suggestions. #chatgpt #AI

13

2 Comments

Big Samsung event this week in AI: Samsung x Google XR device confirmed for later this year, Apple & Microsoft drop out from OpenAI board + much more! Here's what you may have missed: 1. Samsung x Google XR Device Confirmed For Later This Year 2. Samsung Galaxy AI Event Announcements 3. Apple and Microsoft Drop Out From OpenAI Board 4. OpenAI Blocks Access To China 5. Stability AI Releases Stable Assistant Features 6. Fryean: A Free Open Source Background Remover Released 7. Youtube Adds AI to Remove Copyright Music Allowing Creators to Keep Videos Monetized 8. Meta Unveils MobileLLM an Open-Source LLM for Mobile Phones 9. Artifacts Made with Claude Can Now Be Published and Shared ~~ Follow me to stay updated with the biggest announcements in AI. To see this week's AI Funding news, check out my business partner Sahil Uttarwar's most recent post. What was your favorite story that happened this week?👇🏻

86

107 Comments

🚀 Create voice-activated AI assistant all without entering any line of code 🗣️💬 One of my favourite No-code YouTube channel - Planet No Code just released a new tutorial on integrating Vapi with Bubble to create voice-activated AI assistants - no coding required! 🔑 Key Features: ✅ Build AI assistants with voice interaction ✅ Save and store complete chat transcripts ✅ Easy setup using VAPI plugin in Bubble I can build various AI functions for you base on your business requirements. Let’s chat! Follow me as I share more cool stuff built without having to learn coding! #NoCode #AI #VoiceAssistant #Bubble #SaaS #TechInnovation Video link: https://lnkd.in/g8UsgSer

6

Very interesting! The post highlights the role of AI agents functioning as judges and underscores the necessity for enhanced evaluation methods. It references research from Meta AI, emphasizing the critical need to align evaluation techniques with the capabilities of AI. #aiagents #genai #llms

7

Most AI image generators are just renting you access to your own visual language. What happens when you want to own it? Kent Keirsey speaks on this pretty complex relationship and how Invoke is hoping to change that. Read his full interview here: https://lnkd.in/eTq2dUvr #artificialintelligence

18

I created a 5-part guide on how you can — build your 'AI Expert' brand. The demand is through the roof. And most industries don't have their 'go-to-expert' yet. Will it be you? First mover advantage here is key. Plus I'll show you how I monetise. Want it free? Send me a connection. + Comment 'brand' below. And I'll ping it over.

3

1 Comment

🚀 Breaking News: Google's Gemma 2 Shakes Up the AI Landscape! 🌟 Two sizes: 9B and 27B parameters Trained on a staggering 13 trillion tokens Open-source, democratizing access to cutting-edge AI But here's the kicker: Gemma 2 27B is giving Meta's Llama 3 70B a run for its money! 🏃‍♂️💨 Early Chatbot Arena evaluations place Gemma 2 27B in the same league as: Anthropic's Claude 3 Sonnet Meta's Llama 3 70B OpenAI's GPT-4 What does this mean for the AI community? More efficient models = lower computational costs Increased accessibility for researchers and developers Potential for faster iteration and innovation The AI race is heating up, and it's thrilling to see open-source models pushing the boundaries of what's possible. 🔥 What are your thoughts on Gemma 2? How do you think this will impact the AI landscape? Let's discuss in the comments! 👇 Technical Report: https://lnkd.in/gzp3kXFj #AI #MachineLearning #Gemma2 #Google #OpenSourceAI #TechInnovation

2

🚀 **Introducing Gemma 2: Unleashing Next-Gen AI Performance!** 🚀 Gemma 2 offers best-in-class performance, running at incredible speed across different hardware and seamlessly integrating with other AI tools. AI has the potential to address some of humanity's most pressing problems — but only if everyone has the tools to build with it. That's why earlier this year we introduced Gemma, a family of lightweight, state-of-the-art open models built from the same research and technology used to create the Gemini models. 🌟 We’ve continued to grow the Gemma family with **CodeGemma**, **RecurrentGemma**, and **PaliGemma** —each offering unique capabilities for different AI tasks and easily accessible through integrations with partners like Hugging Face, NVIDIA, and Ollama. 🤝 Now we’re officially releasing **Gemma 2** to researchers and developers globally. 🌍 Available in both **9 billion (9B)** and **27 billion (27B)** parameter sizes, Gemma 2 is higher-performing and more efficient at inference than the first generation, with significant safety advancements built in. In fact, at 27B, it offers competitive alternatives to models more than twice its size, delivering the kind of performance that was only possible with proprietary models as recently as December. And that’s now achievable on a single NVIDIA H100 Tensor Core GPU or TPU host, significantly reducing deployment costs. 💡 ### Key Highlights: 🔥 **Outsized Performance:** At 27B, Gemma 2 delivers the best performance for its size class, and even offers competitive alternatives to models more than twice its size. The 9B Gemma 2 model also delivers class-leading performance, outperforming Llama 3 8B and other open models in its size category. 📈 💰 **Unmatched Efficiency and Cost Savings:** The 27B Gemma 2 model is designed to run inference efficiently at full precision on a single Google Cloud TPU host, NVIDIA A100 80GB Tensor Core GPU, or NVIDIA H100 Tensor Core GPU, significantly reducing costs while maintaining high performance. This allows for more accessible and budget-friendly AI deployments. 🏷️ ⚡ **Blazing Fast Inference Across Hardware:** Gemma 2 is optimized to run at incredible speed across a range of hardware, from powerful gaming laptops and high-end desktops, to cloud-based setups. Try Gemma 2 at full precision in Google AI Studio, unlock local performance with the quantized version with Gemma.cpp on your CPU, or try it on your home computer with an NVIDIA RTX or GeForce RTX via Hugging Face Transformers. 🖥️ For detailed performance breakdowns, check out the technical report. 📊 #AI #MachineLearning #Gemma2 #Innovation #Technology #Performance #CostEfficiency #HuggingFace #NVIDIA #CloudComputing #Research #Developers #AICommunity ---

2

ChatGPT Search (née SearchGPT) is less than a month old, but it's a big change in how the chatbot works. In this week's practical advice on how journalists can better use AI, Christopher Allbritton walks through the subtleties of OpenAI's AI-powered search engine. One thing I learned from his excellent piece: How to prompt ChatGPT so it uses current news to suggest interview questions. Paid subscribers to The Media Copilot can check it out all the tips here 👉 https://lnkd.in/eE_S8Q6Q #AIjournalism #media #AIsearch

4

My tuppence worth on 'OpenAI o1', the company's new reasoning model: ➡️ we are now entering a world of multiple state of the art models per vendor - traditionally the likes of OpenAI and Anthropic have had their flagship model, now OpenAI has two: GPT4o and OpenAI o1. They admit that for certain use cases (e.g. copy writing) you'd still use GPT4o and they go on to say: "We also plan to continue developing and releasing models in our GPT series, in addition to the new OpenAI o1 series." So we will have a successor to GPT4o as well as an OpenAI o2. The world of model testing and consultancy (disregarding useless benchmarks), is well and truly upon us. ➡️ given the rumours that OpenAI might be working on a model that costs thousands of pounds to use, that introduces an interesting business opportunity. How much would your business pay for temporary access to a model that could review and develop your strategy or run your board meetings every quarter? It's akin to hiring the very best consultants... OpenAI o1 is not that expensive but it is three times as expensive to enter info into ($15 vs $5 per million tokens) and four times as expensive to get info out of ($60 vs $15 per million tokens) than GPT4o. We're likely approaching a point where companies will need to limit access to more expensive models to select teams/individuals/projects. ➡️ "The o1 models introduce reasoning tokens. The models use these reasoning tokens to 'think', breaking down their understanding of the prompt and considering multiple approaches to generating a response... While reasoning tokens are not visible via the API, they still occupy space in the model's context window and are billed as output tokens...this includes both the invisible reasoning tokens and the visible completion tokens. The maximum output token limits are: o1-preview: up to 32,768 tokens" - this is all taken from the developer documentation. Basically OpenAI's saying that its AI can work out pHD level problems in ~24,576 words or less by reasoning with itself, outputting its thoughts, picking a chain of thought to pursue, and then disregarding everything else over multiple turns (see image below). It's also another reason this new model shouldn't be used for tasks where you need to write long prompts. It uses a lot of the context window you would traditionally use for few shot, for thinking. Means reasoning can be scaled with longer context windows... ➡️ "...these models are not supported in the Assistants API" - this is the bit of tech that powers our internal agents at Definition and many apps around the world. I'm really excited about the potential of powering Assistants with OpenAI o1 - they will become much more capable and able to handle far more complex use cases e.g. extensive strategy planning. ➡️ you know how commentators like to point out LLMs don't 'think' ("...they're prediction machines; predicting the next token isn't thinking..."), are we now at a point where they are 'thinking'?

5