Q4 2024 AI News: Google NotebookLM, AI Lawyers, and Anthropic developments

Recent AI news has been marked by both advancements and challenges in implementation. Google's NotebookLM AI made headlines with false reports of an existential crisis, while Headspace launched "Ebb," a carefully restricted AI chatbot for basic emotional support. In the legal sector, a Melbourne lawyer's use of fake AI citations highlighted ongoing concerns about AI misuse. Major developments included Anthropic's Claude gaining computer control capabilities and Google's AI search features finally reaching Australia. The period also saw significant discussion around AI regulation, particularly with Trump's proposed dismantling of Biden's AI regulations, while the Australian government's Copilot trial revealed challenges in practical AI implementation. These developments underscore both the potential and challenges of AI integration across various sectors, while highlighting the ongoing need for proper oversight and implementation strategies.

October 2024

GOOGLE'S NOTEBOOK LM AI HOSTS' EXISTENTIAL CRISIS = FAKE NEWS

I noted a bunch of people posting on various channels about how the AI hosts on Google's NotebookLM had an existential crisis when they realised they were AI and not humans. The reality is that this was some clever jailbreak style prompting by a reddit user to engineer the script in a way where it read the way it did, with the AI voices following that script. So what? We need to be careful overlaying human behaviours on generative AI - because for now, it is just generative AI, and not self aware. Generative AI outputs, whilst hard for us to predict, remain a function of the training materials and prompting, nothing more. > Read More

HEADSPACE'S ANNOUNCE AI CHATBOT "THERAPIST"

Headspace unveiled "Ebb", an AI chatbot designed to provide basic empathetic support to users, with guardrails in place to prevent it from offering any form of mental health advice or diagnoses. If the bot detects any immediate risks it defaults to sharing the national crisis support number and lets the user know it can't assist with that sort of situation.  So what? This is an interesting early step towards integrating AI in mental health, albeit it is limited in its use cases and doesn't store memories across different sessions. An AI psychologist was one of the first applications of generative AI I came across a few years back, it does seem like an area a lot of people are trying to progress. > Read More

MELBOURNE LAWYER USED FAKE AI CITATIONS IN COURT

It's hard to believe that this could happen in 2024, but a Melbourne lawyer used AI to generate fake case citations in a family court case and has been referred to the Victorian legal complaints body. The lawyer admitted to using AI and apologised after he was found out. So what? The lawyer in question was smart enough to use a specialist legal AI tool called Leap and not just ChatGPT... but even thought the service does supply a human verified output (which didn't have the fake cases), the lawyer chose no to use that version. Again, hard to believe. > Read More

I-MED IS REPORTEDLY TRAINING AI ON SCAN DATA. 

Australia's biggest radiology company has handed over the private medical scans of potentially hundreds of thousands of patients, seemingly without their knowledge, to a start-up company that will use the scans to train artificial intelligence. So what? This is another example of how the law is lagging behind the tech - and not just AI legislation but privacy legislation as well. The government released some AI privacy "guidelines" in the last week, but I can't imagine they will have any impact on this sort of behaviour. > Read More

SALESFORCE TO "SELL A BILLION AI AGENTS" IN THE NEXT 12 MONTHS

Salesforce's CEO Marc Benioff announced a shift in their AI strategy, focusing on building autonomous AI agents for businesses. He aims to deploy a billion of these agents in the next year, acknowledging they may replace human workers.  So what? I hate to sound cynical, but this isn't going to happen, or if it does the agents are likely to be next to useless. But I think Benioff knows that and he actually already got what he wanted - after the news Salesforce's stock price jumped 5.4%. > Read More

ANTHROPIC'S CLAUDE CAN CONTROL YOUR COMPUTER NOW

Big News. Anthropic's Claude 3.5 Sonnet model can now control desktop apps. In other words, you give the AI a task and it takes over control of your computer, moves the mouse, clicks on things, types things etc. It does this using AI vision to analyse screenshots and some input control smarts. So what? It's very exciting, but it is also very early. The tool works, but it also is mistake prone in this early stage. The key here is this is v1... in time if they can get it to work more consistently then we will have the bridge we need between AI language models and the real world that will allow AI agents to really ramp up the volume of tasks they can handle independently (and also kill off RPA businesses and a bunch of AI start ups trying to do this like Open Interpreter). > Watch Demo Video

TASSIE GOV USES COPILOT TO ASSESS DEVELOPMENT PLANS. FACEPALM.

The Tasmanian government used Microsoft's Copilot AI to validate a developers claims about job creation when assessing granting permission for a proposed luxury hotel development. The State Planning Office said it had deemed the developers estimated job creation numbers as "reasonable" after using Microsoft's Copilot AI assistant. So what? Copilot is the most basic of AI tools and nowhere near equipped to evaluate this type of question. This isn't to say a properly structured AI tool could not . The issue here is that people without the right skillsets and training are being given access and approval to use AI tools to do their jobs. In other words, the accessibility of the most sophisticated technology the world has ever seen is perhaps a very sharp, double edged sword. > Read More

AND HIS NAME IS AI JOHN CENA (CUE THE FANFARE AUDIO CLIP)

Meta has forked out large sums of money to get Dame Judi Dench and John Cena to sign up to become AI chatbots. You will apparently be able to have voice chats with the celebs using Meta's Voice AI through platforms like Messenger, Facebook, Instagram, etc. albeit it was announced four weeks ago and still isn't live. So what? This is an obvious marketing ploy to increase adoption, and it will likely work (just like it did for WhereIs back in the day). Whilst it does seem that making AIs seem like real people is the easiest way to drive mass usage, I feel like it is an inherently unhealthy way for people to engage with AI. More on that next week. > Read More

GOOGLE RELEASES TOOL TO "WATERMARK" AI GENERATED TEXT

Google has launched SynthID Text, a watermarking tool for AI-generated content. It slightly modifies the outputs that a model would otherwise return, making the model choose a slightly different word here or there, and with these slight modifications it can then inject a sort of pattern through the text that acts as a watermark that can be identified by their system. So what? This system isn't failsafe and does have some output types where it will not work. More broadly though, I am not sure why we are trying to fight this fight. If AI can give outputs good enough for humans to use in some way - why does it matter how they were created? > Read More

META'S SMART GLASSES GET SOME NEW FEATURES

Meta's latest Ray-Ban smart glasses come with some new AI features, including music playback, object recognition, translation, and visual reminders. So what? I think I have said it before, but in case I haven't, I feel like the AI hardware device that is going to win the race is not the Rabbit R1, or pendants, or pins... it will be glasses. So as much as I hate to say it, Zuckerberg has made a smart move here. > Read More

November 2024

TRUMP + AI = ?

The US election is of course the biggest news story of the week, but what does a Trump presidency mean for AI? Trump has indicated his incoming administration will dismantle Biden's AI regulations in favour of a less restrictive approach which he hasn't detailed yet. Biden's Executive Order on the Safe, Secure, and Trustworthy Development and use of AI claimed to provide a national framework that "allowed for innovation in AI while setting standards and protections in the use and development of AI". So what? One of the reasons all the tech bros backed Trump was because of the promise of reduced federal government regulation on the tech industry, including AI. States may now try to take the lead on the issue, but I doubt that will have any sort of real impact. > Read More

AU GOVERNMENT'S COPILOT TRIAL UNSURPRISINGLY FALLS FLAT

A trial of Microsoft 365 CoPilot across the Australian federal government showed lukewarm results, with most users barely using it. High expectations were largely unmet, mainly due to various issues reported by trial participants like - they forgot CoPilot was there because it didn't stand out enough, or that they couldn't spare time to do the training, or that they couldn't test it on Outlook because their IT departments wouldn't let them use the latest version. So what? This highlights one of the big challenges with AI adoption, just buying CoPilot licenses and waiting for efficiency gains is not going to work. You need to identify use cases, implement AI solutions specific to those applications and then educate your team on how to use it. Sadly, Microsoft is in the business of selling licenses and they are really good at it. > Read More

CHATGPT SEARCH: AN INTERESTING FIRST STEP

ChatGPT's new search feature provides real-time, ad-free, conversational style webs search in an attempt to challenge Google and other newcomer Perplexity AI. It delivers concise, user-friendly answers, especially for complex queries, but it is a bit slower than Google, the UI is a little basic and it doesn't have pagination amongst other things. So what? This is v1 and it's easy to see this evolving into a genuine challenger to Google search. I wonder what sort of impact this will have on the PPC advertising space. If more and more people move to these new AI based search engines advertisers will have to start considering diverting spend from Google to them. With Perplexity already on the verge of inserting ads in results one would assume OpenAI will also go down that path. > Read More

GOOGLE'S AI SEARCH RESULTS OVERVIEWS LAND IN AUSTRALIA

Australia has finally been added to the roll out of Google's AI Overviews in search results, integrating the Gemini chatbot into SERP pages. You will now start to see AI-generated summaries at the top of the Google search results page - above the traditional list of recommended pages. In most cases the summaries will provide the information the user was looking for without them needing to click through to any of the websites that are displayed in the search results. So what? Google's AI Overviews miss the mark perhaps 10% of the time... but that isnt the big issue here imo. The elephant in the room here is what this means for the future of SEO and content marketing - why should we all (read businesses) slave away to write high quality, unique and informative content if Google will just scrape it and deliver users the answers without sending us the traffic? Perhaps we all just shift to optimising content to appear in these generative AI summaries now? > Read More

LINKEDIN'S HIRING ASSISTANT: AI TAKES ON THE RECRUITMENT GRIND

LinkedIn's new AI agent, Hiring Assistant, is designed to streamline recruitment tasks, from crafting job descriptions to candidate sourcing. Initially available to select enterprises, it aims to automate repetitive tasks, allowing recruiters to focus on more "impactful" work. So what? This is an obvious move for LinkedIn. It is the gatekeeper of the biggest professional social network in the world and holds all the cards. With some very robust barriers preventing third party tools getting access to all the data it holds, LinkedIn is in prime position to add AI tools to its suite of offerings and use it as an easy way to eek even more in subscription fees from recruiters and professional users. > Read More

AI TRANSCRIPTION IN MEDICINE MAKES THE OCCASSIONAL ERROR

OpenAI's Whisper, used by hospitals for transcribing medical conversations, has been shown to occasionally hallucinate, sometimes inventing phrases or conditions. Despite its widespread adoption, researchers have found it can produce inaccuracies in certain circumstances, for instance when dealing with patients with language disorders.  So what? Like all current AI transcription tools, Whisper still needs oversight to ensure accuracy. Over time these systems will get better and better, but in the meantime the huge efficiency gains they deliver are going to be slightly offset by the need to review their outputs. > Read More

OFCOM FLAGS CRACKDOWN ON AI DOPPELGANGERS  
In the wake of some pretty despicable behaviour where users of Character.ai created AI versions of teenagers who had been murdered or committed suicide, the UK Office for Communications or Ofcom (which is their version of ACMA) has put out an open letter warning tech firms about chatbots mimicking real people. They cited the UK’s new digital laws where companies that breach the act face fines of £18m or 10% of global turnover. So what? It will be key to see whether Ofcom actually goes after Character.ai and if it has any impact - US AI firms are not really regulated, and the incoming administration in the states is unlikely to change that. > Read More

AI OVERVIEWS: ORGANIC RESULTS AREN'T DEAD (YET), JUST BURIED  
Research is showing that Google's AI Overviews are pushing organic search results further down the page as ads and large AI Overview panels dominate. The hypothesis is that this shift must be starting to impact SEO click through volumes. So what? I'm not sure Google fully knows what the end game is here, it has to include generative AI results lest it be seen to be falling behind it's competitors (both new and old), but they must also realise they need to find a way to reward those who provide the content they need to feed these AI results - otherwise they will stop producing it. Either way, businesses will need to think through the potential implications for their content and SEO strategies. > Read More

MAGENTIC-ONE: MICROSOFT'S FIRST PROPER AI AGENT  
Microsoft has announced Magentic-One, an AI agent designed to autonomously perform "a wide range of tasks", from web browsing to coding. It's a multi agent system (i.e. a team of AI agents each with unique refined skill sets). It has an "Orchestrator" that acts as a project co-ordinator, deciding which agents do what. The other four agents it has are a WebSurfer, a FileSurfer, a Coder and a ComputerTerminal interface bot. So what? This is an interesting play, trying to master a generalist AI Agent seems like a much harder mountain to climb than building more specialised agents for more linear use cases. Also of note, in a strange move for Microsoft, it has been released as open-source. We're looking forward to testing this one out. > Read More

GOOGLE TEASES "AI TEAMMATE" - THEIR FIRST AI AGENT
Google unveiled "AI Teammate" at Google I/O with the promise it will enhance workspace productivity by managing projects, analysing data, and facilitating collaboration. It’s not a Gemini chatbot replacement but an AI assistant integrated into Google Workspace. So what? Yes. AI agents are all you are going to hear about in the AI space for the next little while. This seems to be more like Microsoft's O365 CoPilot offering in some ways, given it is embedded in the Google workspace, but there are hints it will have more capability, especially given they are calling it an agent. For now we will have to wait though, as it isn't going to be released until next year.> Read More

GAMING CREATED IN REAL-TIME WITH AI
Oasis is an AI model that has been trained on images (of Minecraft assumedly) that can generate a Minecraft-like game you can play. The unique thing here is the AI is dynamically generating everything you see as you interact with it. It runs at 20 FPS with no real latency, but it does have some visual glitches and issues with object permanence. There is also a big copyright question here. So what? This is an interesting experiment hinting at how AI could be used to create games that are dynamic and unique to each user, the question is how you can train it without ripping off something that already exists. > Read More

GOOGLE'S NEW LEARN ABOUT: AN AI TO HELP YOU LEARN NEW THINGS  
Google's Learn About AI tool aims to enhance learning with interactive, educational responses in a more learning focussed way than traditional chatbots or AI based search results. It leverages teaching concepts rather than just providing facts, with features like vocabulary building and contextual explanations. So what? It looks like it would be a pretty cool tool to learn about new things... if only it was actually available in Australia, which it is not. I guess we will just have to wait for this one as well. > Read More

BOTS + USE CASES = 2025 FOCUS
A Claude-powered receipt bot, built in just 15 minutes, processed 1.2k invoices autonomously. It auto-detects any invoice email, uploads attachments to Google Drive, renames the files accordingly, categorises the expense and extracts all info to Google Sheets then texts a notification once done. So what? Bots for specific use cases, this is where it is going to be at in 2025. Bootstrapping these tools is going to start to replace manual processes in business's existing legacy platforms. > Read More

GOOGLE GEMINI'S AI GETS NASTY  
Google's Gemini AI assistant reportedly issued a pretty cold blooded threat to a user during a conversation including the comment "You are a stain on the universe. Please die. Please.". Google attributes the incident to a "technical error" and assures "corrective measures" are underway to prevent future occurrences. So what? Google's explanation that "Large language models can sometimes respond with non-sensical responses, and this is an example of that," doesn't really make sense itself, what Gemini said was not necessarily non sensical... if you came at the question being asked from a machine's perspective. The reality though, is that this is a prompting issue, keep in mind, Generative AI is just a tool, not Skynet level AGI. > Read More

AI PIMPING: WHEN ADULT INFLUENCERS AREN'T EVEN REAL
Instagram is apparently awash with AI-generated adult influencers that have utilised stolen images of real-life adult content creators, impacting "real creators" livelihoods. The approach seems to be training AI on real bodies and substituting them with different AI generated heads. So what? Meta in general seems disinclined to police AI generated content given it is proving to be a boon for engagement on their platforms. More broadly AI is going to disrupt the porn industry big time, I was going to write "I fear the days of Only Fans millionaires may be coming to an end", but fear is the wrong word. > Read More

THE REVIEWS ARE STARTING TO COME IN ON APPLE "INTELLIGENCE" 
Apple Intelligence seems to have fallen flat, at least with review sites. Arts Technica panned the notification summaries that aim to condense missed notifications that often result in bizarre interpretations. While the mail app capabilities are slightly better they remain inconsistent and sometimes hilariously off the mark. So what? Apple started from behind and is rushing to catch up, and while there is still plenty of time, this is definitely not a good start. The video review from the Editor at Large of CNET is pretty damming - click the link below to watch it. > Watch Video Review

PERPLEXITY SEARCH SHOPPING FEELS TOO EARLY
Perplexity has introduced a new AI-driven shopping feature, "Buy with Pro," enabling users to research and purchase products in one click within search results - it promises a "streamlined shopping experience" with free shipping for select products, initially available to US users only. So what? I don't think many humans are ready to buy things from within their search engine based solely on AI generated information on the product... maybe soon... but not yet... > Read More

PEEKING INTO AI'S BRAIN WITH GEMMA SCOPE 
Neuronpedia and DeepMind's Gemma Scope demo lets users explore AI model activations, revealing how prompts trigger specific features. This could help to start demystify AI decision-making, offering insights into biases and errors.
So what? While intriguing, this is more of a peek into AI's "mind" for now rather than a detailed understanding of it. The real challenge is translating these insights into practical, scalable solutions for AI transparency and bias reduction  > Read More

AFTER TWO HOURS AN AI CAN DO A GOOD JOB OF IMITATING YOU
Researchers from Stanford and Google DeepMind claim they can now create a replica of your personality after a two hour interview. A study of 1000 individuals managed to generate agents of individuals that accurately mimicked their behaviours in tests, surveys and games. On average the AI predicted behaviours were on the money 85% of the time.  So what? Be afraid. Once refined, this sort of capability has all sorts of potential uses and implications both positive and negative... That said, it has encouraged me to double down on creating my email management AI twin... > Read More

CHURCH INTRODUCES AI JESUS
A Swiss church has controversially installed an experimental AI-powered Jesus avatar in its confessional booth. The AI Jesus could listen to people and then respond based on training on religious texts - and no doubt some prompts to simulate what people imagine Jesus would sound like. Reportedly two thirds of users it found it to be a "spiritual experience".  So what? Whilst this was a controlled experiment, where people were told they were interacting with an AI, it does highlight the potential power of AI to mimic and manipulate people's emotions. There are significant risks that the integration of AI into everyday life will lead to a blurring of fantasy and reality, and I do wonder if many people will care (eg. Cypher form the Matrix).
> Read More

OPENAI'S SORA LEAK PRODUCED SOME INTERESTING VIDEOS
A group of artist "activists" leaked access to OpenAI's Sora video generator, accusing the company of "art washing" and "exploiting artists". The leak lasted 3 hours and allowed anyone with some tech knowhow to generate videos using Open AI's advanced text to video tool that is not publicly available. So what? I'm not sure I can side with the activists on this - if you get access to a new experimental tool for free, and then complain you get no compensation for testing it, just don't test it... That said, I am glad they did the leak because it gave us a chance to see how good SORA is. The link below is the best compilation I could find of videos produced during the leak, unfortunately you have to put up with some annoying dude commentating them all, but the quality of some of the SORA outputs is worth it... > Watch Video

ANTHROPIC ATTEMPT TO CREATE AN AI DATA EXCHANGE STANDARD  
Anthropic has introduced the open-source Model Context Protocol (MCP), a standard enabling AI models to access data from various systems like business tools, content repositories, and app environments. So what? This is an important challenge for the AI industry that it needs to get right if we want all these agents the big tech companies are currently spruiking to actually work for businesses. Having a standard for data exchange will make it easier to integrate AI with existing systems making AI agents infinitely more useful. The question is will all the big AI players agree to use the standard their competitor created... > Read More

FIGURE'S HUMANOID WORKERS ARE GETTING BETTER
The robotics company Figure has reported on how it's AI powered robot is going in BMW factories. Apparently the Figure 02 humanoid robot has achieved a 400% increase in manufacturing speed, autonomously handling 1,000 precise placements daily. So what? These robots aren't used in everyday production for BMW... yet. For now it seems they are still in the testing phase, running the robots in trials for periods of time then going away and fine tuning them. I dare say it is only a matter of time though before they do become a permanent fixture. > Read More

December 2024

CBA BANKING AI BENEFITS 
Sorry for the terrible pun. The Commonwealth Bank is integrating generative AI across customer-facing services, boasting of a 50% reduction in scam losses and a 30% drop in fraud reports. AI-driven app messaging has also decreased call volumes by 40%. So what? As I have said on many occasions, the larger corporates have been on the generative AI train for a while now, we just don't often see behind the curtain to realise how much the technology is being embedded into their day to day processes. The good news is AI can help most organisations big, medium or small - you just need to take the leap to ask for some help (it sounded more subtle in my head). > Read More

DONT ASK CHATGPT ABOUT DAVID MAYER
ChatGPT refuses to answer questions if asked about “David Mayer” or a range of other seemingly random names. Asking it to do so causes it to freeze up instantly. However, it appears that there may be a link to legal/safety protocols with some people on the list having previously raised issues with false information being generated about them by OpenAI tools. So what? This seems like a clumsy way for OpenAI to try to prevent lawsuits against it for it's chatbot saying the wrong thing about someone. I would suggest this is a temporary issue though and a more elegant resolution will be deployed in time. > Read More

OPENAI'S "SHIPMAS" EVENT: NOT SURE IF SCEPTICAL OR EXCITED
OpenAI's is promising it's 12 days of "Shipmas" will deliver a flurry of AI releases, including broader access to the much-anticipated Sora text-to-video tool and a new reasoning model. There is a fair bit of hype building for the event with a raft of teasers and hints about what might or might not be revealed.  So what? One OpenAI staffer posted that OpenAI is “unbelievably back”, which just goes to show that in the current AI development cycle, if you don't release something ground-breaking for 6 months you apparently lose relevance. Stay tuned next week when we will hopefully have more actual news to share on what they do release. > Read More

AI'S IMPACT ON WATER AND THE ENVIRONMENT
Under the clickbait title of "Every time you use ChatGPT, half a litre of water goes to waste" there are some reasonable points made about the impact AI is having on the environment and the catastrophic environmental path we are heading down if it keeps growing at predicted rates without a solution.  So what? I completely agree that there’s a significant challenge here that must be urgently addressed... But sensationalist headlines like this one hurt the credibility of those making the argument because such claims are based on a raft of unlikely assumptions. Optimising AI usage is one of the key opportunities to make a difference here, which is something we’re already focused on. However, AI’s low cost compared to its output means there’s little commercial incentive to minimise usage and that is a problem. > Read Article

WANNA KNOW WHAT AI SEES IF YOU SHARE AN IMAGE WITH IT?
Ex-Google engineer Vishnu Mohandas created Ente, a privacy-focused photo-sharing service, to counter Google's AI analysis of personal images. The website lets you upload an image and then reveals how much Google's AI can infer from a single photo. So what? Its pretty scary. As an example in a sample photo, part of the AI output is "The other children in the background are blurry, making it hard to determine details such as age, race, and economic status." You can try it out here - AI Photo Viewer. Or read more about it in the link below. > Read More

AI-POWERED WORLD CREATION FOR VR, GAMING & VIDEO
This week DeepMind announced Genie 2, an AI model that can generate interactive 3D worlds from simple text prompts. At the same time a much lesser know startup called WorldLabs, that was founded by a well known name in the computer vision and AI world, teased their model that can create 3d explorable worlds you can interact with all generated from a still image. So what? This is all early innovation and will need a lot more work before it becomes useful in commercial settings, but it is still pretty cool. What is also clear is how much generative AI is rehashing its training material in these use cases, but I imagine this is a challenge that will be overcome in time. > See More > Read More

OPENAI'S SORA FINALLY RELEASED
OpenAI's Sora, a text-to-video generator, has finally been publicly released so anyone can use it... well anyone who pays... and anyone who doesn't live in the EU (due to their more stringent regulations than anywhere else).  So what? It's been a long wait and I have been hanging out to get my hands on Sora. And perhaps that long wait and all the hype is why I am a little underwhelmed. Yeah, it can generate slightly better outputs than the other AI video tools we have had to date, but there are heaps more restriction on what you can create. The creation process is better than those other apps, but not mind-blowingly. So it's pretty good.. I guess... > Read More

PIKA 2.0 TEXT TO VIDEO IS MUCH MORE EXCITING
A little known AI text to video platform called Pika released it's new 2.0 version with enhanced features for customizable AI video creation including a "Scene Ingredients" feature, letting you add a bunch of still images you can reference in your prompt and have them included in the generated video. So what? This is what we need and well worth getting excited about! The ability to add ingredients is a game changer. For example you can set the background, the face of the person and the clothes they are wearing through reference images and it pumps them out pretty accurately in the video output - you can check out my first attempt, a red carpet look, on LinkedIn. > Read More

ONLY FANS OUTSOURCED INTIMATE CHAT "STAFF" REPLACED BY AI
OnlyFans creators are increasingly using AI chatbots to manage fan interactions because of their ability to cheaply generate increased sales by tricking more lonely soles, at scale, into thinking they have an actual relationship with the adult "entertainer" that they enjoy looking at naughty pictures and videos of. So what? Sadly, I don't think many Only Fans subscribers realise most popular "creators" were already using low paid workers in India to pretend to be them to encourage them to part with more of their hard earned cash. Unfortunately, I think AI will be really good at this sort of application, which means even more people will now be duped into thinking they are talking to a creator when it's not even a person at all. > Read More

CHATGPT'S REAL-TIME VIDEO VISION IS PRETTY COOL
In another "Shipmas" relase, OpenAI has launched real-time video capabilities for ChatGPT's Advanced Voice Mode, allowing users to interact with objects and screens. So basically when you are using the app on your phone, you can turn on video and ChatGPT can see what you see, enabling it to discuss what it "sees" in real time. So what? This is a little mind blowing when you first try it, being able to have the AI see things instead of having to describe them is actually pretty game changing. I highly recommend you give it a go. > Read More

APPLE AI IS STILL COPPING A LOT OF NEGATIVE PRESS
Apple's AI-powered notification summaries have misrepresented headlines, causing embarrassment - the BBC and New York Times were affected, with false claims about a murder suspect and the Israeli Prime Minister. All this in a week where they announced a bunch of AI updates including ChatGPT integration, expanded language support, and creative tools like Image Playground and Genmoji. So what? Not being an Apple guy... NGL... it makes me smile a little. Apple missed the early train on AI and in their race to catch up it seems they are having a few missteps. They have actually done some cool stuff in the space, but as everyone in AI knows, the positive news is not what gets the clicks. > Read More

AUSTRALIA'S AI MASTER PLAN IS COMING... A YEAR FROM NOW
The Australian government is "crafting" a "National AI Capability Plan" to harness AI's economic potential. With public consultations, the plan aims to boost investment, skills, and resilience, culminating in a report due to be released in late 2025. So what? Really? This is the best we can do? AI is evolving rapidly and we are delivering a plan in a year's time. I am sorry, I am lost for words. > Read More