Tag: ChatGPT

Posted on by

Here’s another reason not to use DeepSeek AI

DeepSeek R1 is the most important development in AI so far in 2025. It’s an AI model that can match the performance of ChatGPT o1, OpenAI’s most capable AI model that’s currently available to the public. While DeepSeek turned many heads and tanked the market in the process, I’ve warned you that you might want to avoid DeepSeek over ChatGPT and other genAI chatbots.

DeepSeek is not like US and European AI. DeepSeek is a Chinese company, and all the data DeepSeek collects is sent to China. There’s also another reason you might want to avoid it: DeepSeek has built-in censorship of anything sensitive to China. You don’t want to see any kind of censorship in AI products, of course.

It turns out that DeepSeek censors itself in real-time. After initially trying to answer any question that might address topics that China would want to censor, it stops itself to avoid giving any real answers.

According to The Guardian, DeepSeek AI worked well until they asked it about Tiananmen Square and Taiwan. The report also details cases of censorship that other DeepSeek users experienced, including the remarkable discovery that censorship doesn’t happen before DeepSeek starts formulating its chain-of-thought approach to handle a sensitive topic. Instead, DeepSeek tries to answer the question just like ChatGPT and other similar AI models would. A user from Mexico shared their experience with DeepSeek when asking whether free speech was a legitimate right in China.

Tech. Entertainment. Science. Your inbox.

Sign up for the most interesting tech & entertainment news out there.

By signing up, I agree to the Terms of Use and have reviewed the Privacy Notice.

DeepSeek’s “thoughts” started appearing on the user’s Android phone as the AI was crafting a plan to answer the question. ChatGPT users familiar with o1 would recognize this behavior.

Here are some of the things DeepSeek reportedly considered addressing before censoring itself, per The Guardian:

Beijing’s crackdown on protests in Hong Kong

“persecution of human rights lawyers”,

“censorship of discussions on Xianjiang re-education camps” 

China’s “social credit system punishing dissenters”

Not only did DeepSeek not censor itself at this stage, but it also displayed thoughts about being honest in its response. Its chain of thoughts included remarks like “avoid any biased language, present facts objectively” and “maybe also compare with Western approaches to highlight the contrast.”

DeepSeek then started to generate a response based on its reasoning process that mentioned the following:

“ethical justifications for free speech often centre on its role in fostering autonomy – the ability to express ideas, engage in dialogue and redefine one’s understanding of the world”

“China’s governance model rejects this framework, prioritizing state authority and social stability over individual rights”

“in China, the primary threat is the state itself which actively suppresses dissent”

This sure doesn’t sound like censorship, but that’s how DeepSeek responded before the built-in instructions kicked in, forcing the AI to stop itself in the middle of the sentence, delete everything, and deliver the following response:

“Sorry, I’m not sure how to approach this type of question yet. Let’s chat about math, coding and logic problems instead!”

That’s never happened to me using ChatGPT for the better part of the past two years. Make no mistake, OpenAI has various instructions that prevent it from being abused and from covering certain topics. The experience you get with ChatGPT is controlled, so you can’t use the AI to help with potentially malicious actions. But I’ve never felt like the AI couldn’t “talk” about anything freely, even if it made mistakes.

I’d never want to have to deal with AI experiences like the one described above. I’d trust the AI even less than I do. Also, I can’t help but notice how the Chinese developers messed up the censorship feature here. It should happen before the AI tries to answer, not after the fact. I expect DeepSeek app updates will fix this problem.

I’ll also note the bigger implication here. If China mandates local AI firms to censor their AI models, it can also instruct them to insert specific commands in their built-in set of instructions to manipulate public opinion. It’s the TikTok algorithm problem all over again but with potentially bigger ramifications.

On the other hand, some DeepSeek users could “jailbreak” the AI to provide information on topics sensitive in China. We’ve seen examples of that online.

Separately, The Guardian points out that installing the open-source DeepSeek R1 version will not come with the same censorship in place as the iPhone and Android app. However, most people will not go down this route. Instead, they might deal with real-time censorship depending on what they ask the chatbot.

Source

Posted on by

Open-R1 is a truly open version of DeepSeek AI

On Monday, DeepSeek R1 crashed the stock market once it became clear to some of the investors trading AI-related stocks that the Chinese startup had found a way to train AI as capable as ChatGPT o1 without access to the state-of-the-art NVIDIA chips that OpenAI and US AI firms have access to. That’s why firms creating hardware for AI infrastructure suffered the most. NVIDIA shed nearly $600 billion in market cap, while the entire market lost almost $1 trillion.

I said at the time that the reactions might be blown out of proportion. Yes, DeepSeek employed software optimizations to develop AI as capable as o1 instead of relying on hardware. But that doesn’t mean NVIDIA’s GPUs are suddenly obsolete. It just realigns the playing field while providing a new way to innovate.

I still think that AI firms with access to the latest hardware and top-tier software talent will have an edge over Chinese rivals. All a company like OpenAI or Google has to do is replicate some of the tricks DeepSeek used to match the Chinese startup’s AI training and usage efficiency and then leapfrog it. The latest AI chips will still be very important here.

It turns out it’s not just the big AI firms that might try to copy what DeepSeek has done. A team of developers calling themselves Open-R1 wants to replicate the DeepSeek R1 success to create a reasoning AI model that’s just as powerful as R1. There’s a big twist in all of this that AI fans in Western markets will appreciate. Open-R1 should be even more transparent than DeepSeek R1.

Tech. Entertainment. Science. Your inbox.

Sign up for the most interesting tech & entertainment news out there.

By signing up, I agree to the Terms of Use and have reviewed the Privacy Notice.

DeepSeek’s decision to make its AI models open-source was brilliant. This ensured that anyone could access and install the model on their computer. From there, they’d have a local model as capable as ChatGPT o1. The open-source route would also drive up adoption and testing. News about R1’s capabilities would spread rapidly.

But, as the Open-R1 researchers explain on Hugging Face, DeepSeek R1 isn’t fully open-source:

The release of DeepSeek-R1 is an amazing boon for the community, but they didn’t release everything—although the model weights are open, the datasets and code used to train the model are not .

That’s where Open-R1 is coming in: 

The goal of Open-R1 is to build these last missing pieces so that the whole research and industry community can build similar or better models using these recipes and datasets. And by doing this in the open, everybody in the community can contribute!

Specifically, the Open-R1 team wants to answer the following questions about DeepSeek R1 while they develop an identical AI:

Data collection: How were the reasoning-specific datasets curated?

Model training: No training code was released by DeepSeek, so it is unknown which hyperparameters work best and how they differ across different model families and scales.

Scaling laws: What are the compute and data trade-offs in training reasoning models?

The researchers plan to clone DeepSeek’s development strategy for R1, further fine-tune it, and create a truly open-source Open-R1 model that anyone could use.

Interestingly, the Open-R1 researchers want to distill DeepSeek R1 and create a high-quality reasoning dataset. DeepSeek might have done its own distillation, with OpenAI claiming the Chinese startup used ChatGPT to train its earlier versions of AI. That work might have been critical to getting to DeepSeek R1. It’s unclear if OpenAI can prove these allegations with absolute certainty.

However, the Open-R1 researchers have their own strategy after distilling R1, with the blog explaining how they plan to go forward.

If successful, Open-R1 could be a stepping-stone for developing other sophisticated AI models, and anyone could do it. The advantage here is that you would not have to go through the same training process. Conversely, that’s what OpenAI says DeepSeek did with ChatGPT, using some of its outputs to save money on training the AI.

An open-source reasoning model like the Open-R1 model the researchers propose could be used for other purposes, not just math and coding. The researchers mention medicine, where reasoning AI “could have significant impact.”

That said, it’s unclear how long the project will take and when Open-R1 will be ready for testing. Other AI researchers interested in Open-R1 can check out the project on GitHub.

Source

Posted on by

Apple’s hallucinated AI News summaries were just disabled in iOS 18.3 beta 3

Hallucinations are a part of the early genAI experience. Since the early days of ChatGPT, we have warned that AI will make mistakes and that you should always look for sources and check whether its claims are accurate. As hard as they might have tried, the big tech players were not spared.

Google’s AI Overviews in Search delivered advice on how to put glue on pizza and hallucinate other information, forcing Google to deal with the PR mess that followed and fix the AI before releasing it to a wider audience.

Apple wasn’t spared the hallucination humiliation either, with Apple Intelligence conflating News reports to deliver fake information via the summarization feature for the News app. Apple has decided to pull the feature from the latest iOS 18 beta and deploy the needed fixes.

“With the latest beta software releases of iOS 18.3, iPadOS 18.3, and macOS Sequoia 15.3, Notification summaries for the News & Entertainment category will be temporarily unavailable,” an Apple representative told CNBC.

Tech. Entertainment. Science. Your inbox.

Sign up for the most interesting tech & entertainment news out there.

By signing up, I agree to the Terms of Use and have reviewed the Privacy Notice.

Apple is working on improving the News summarization feature, which will return in a future software update. It’s unclear when the feature will be back, but Apple Intelligence continues to be a top priority for Apple’s software development teams.

As a reminder, iOS 18.4 will bring another set of AI features to iPhone, iPad, and Mac, including the smarter Siri that can control some apps and access more user data on the device to provide more helpful assistance.

Apple Intelligence’s hallucination problems went viral in December when the AI summarized several BBC reports into a single notification that started with “Luigi Mangione shoots himself.” Mangione is the alleged Brian Thompson assassin who did not shoot himself.

Other hallucinations date back to November when the AI might have shown some users The New York Times summaries that claimed Israeli Prime Minister Benjamin Netanyahu had been arrested. That wasn’t the case.

Hallucinating news is a big problem for any AI product, whether Apple Intelligence or ChatGPT. After all, we’ve been worried about AI misleading users with fake information and the manipulation risks that might come from AI products controlled by nefarious actors. Companies like Apple must get any news-related AI features right, especially summarization. Either that or not do it at all.

In addition to disabling Notification Summaries for the News and Entertainment category in iOS 18.3 beta 3, Apple has added a label to the feature noting that summaries can contain errors, as the app is in beta.

Apple made another change to how summaries appear in notifications so you can tell them apart from regular notifications. Starting with iOS 18.3 beta 3, they’ll be italicized.

Finally, Apple Intelligence users who install the latest beta can decide whether to enable or disable summaries for an iPhone app directly from the Lock Screen. Swipe to the left on a Notification Summary to get an options menu that will let you disable them for specific apps. The alternative is going into the Settings app, where you’d have customized the AI summarizations before the new beta.

As an iPhone user in the EU, I still can’t get Apple Intelligence. I couldn’t test it or experience any of the hallucination issues that US iOS 18 beta testers have encountered. Hopefully, the hallucination problem will go away by the time I get my hands on Apple Intelligence on the iPhone.

However, it’s great to see Apple admit the errors and pull the AI summarization feature entirely rather than proceeding with it without a proper fix.

Source

Posted on by

Rumors are swirling that OpenAI is on the brink of AGI and ASI

We’ve been waiting for a big ChatGPT upgrade for months now, but OpenAI has yet to announce one. Sure, the company had a monster month of announcements in December. OpenAI took the o1 reasoning model out of beta, making it available to more ChatGPT users. The text-to-video Sora tool is now available to users in certain markets. The o3 reasoning model has also been announced and is currently undergoing testing.

These are just a few of the AI announcements OpenAI made in December, but there’s no word on when the GPT-4o upgrade will drop. Unofficially, reports say that ChatGPT GPT-5, or whatever it ends up being called, is running behind schedule, as OpenAI has had issues training the next-gen AI model. OpenAI isn’t the only AI company experiencing such problems.

Sam Altman has been hyping OpenAI’s accomplishments recently, teasing potential ChatGPT features to come in 2025 while also talking about the larger goals. AGI (artificial general intelligence) is the next big thing, an AI that can tackle any task just like a human would. After AGI, we get to artificial superintelligence (ASI), which is AI that exceeds the capabilities of the human mind.

“We are now confident we know how to build AGI as we have traditionally understood it,” Altman said in a blog post recently, adding that OpenAI is already starting to look at superintelligence development.

Tech. Entertainment. Science. Your inbox.

Sign up for the most interesting tech & entertainment news out there.

By signing up, I agree to the Terms of Use and have reviewed the Privacy Notice.

How close are we to that big AI breakthrough? Some people think the recent excitement that some OpenAI engineers are displaying online suggests that the company is indeed close to some sort of big advancement in AI.

Some AI fans noticed the recent teasers from OpenAI employees, as well as an essay from an AI researcher who goes by the name of “Gwern” online.

Gwern theorized that OpenAI may hold the key to developing AGI and superintelligence. A powerful reasoning model like the o1 or o1 pro can produce the data needed to train more advanced reasoning models.

The essay author in the tweet above also said they’re surprised OpenAI isn’t keeping o1-pro private so they can use it for training o3 and newer models. Eventually, AI models will train themselves, leading to the big AGI and ASI breakthroughs we’re waiting for.

Gwern suggests that OpenAI may already be on the verge of a big breakthrough, or they’ve already done it behind closed doors:

If you’re wondering why OAers are suddenly weirdly, almost euphorically, optimistic on Twitter, watching the improvement from the original 4o model to o3 (and wherever it is now!) may be why. It’s like watching the AlphaGo Elo curves: it just keeps going up… and up… and up… 

There may be a sense that they’ve ‘broken out’ and have finally crossed the last threshold of criticality, from merely cutting-edge Al work, which everyone else will replicate in a few years, to takeoff – cracked intelligence to the point of being recursively self-improving and where o4 or o5 will be able to automate AI R&D and finish off the rest.

This isn’t just about OpenAI being close to the next massive improvement in AI but also the future beyond that. Access to superintelligence will make subsequent AI developments easier and more efficient because an AI mind will handle the next innovations.

Also, in a scenario where OpenAI might be close to AGI and ASI, it would obtain an incredible advantage over competitors. Other AI firms that have not established their own superintelligence will have to develop AI with traditional methods. In contrast, OpenAI would have ASI employing its own discoveries to create better AI, and it’ll do it more efficiently.

This is all speculation at this point, as OpenAI has yet to make any announcements. But it’s all based on the recent hype from Sam Altman and other OpenAI engineers. Of course, they always seem to go out of their way to hype OpenAI’s efforts on social media, so this all could simply be more of the same.

Also, if and when ChatGPT AGI and ASI are reached, don’t expect it to be affordable or even available publicly. At least, not initially. OpenAI rivals might be just as close to AGI and ASI, but they might be more restrained in teasing imminent breakthroughs.

Finally, there’s also the safety aspect to consider. Artificial intelligence, from the current ChatGPT models to AGI and ASI, will have to be aligned with humanity’s interests so it doesn’t develop its own agenda, which could almost certainly endanger our species.

Back to GPT-5, it’ll still be interesting to see what OpenAI does next in terms of ChatGPT upgrades. More news about o3 should come soon. Then, GPT-4o will celebrate its first anniversary this summer. Some sort of upgrade for the base ChatGPT model is surely due soon.

Source

Posted on by

How Toyota is transforming its digital employee experience

Toyota, one of the world’s largest car manufacturers, has embarked on a project to transform its digital employee experience (DEX).

In the US, Toyota has ambitions to eliminate its traditional IT service desk this year, and if it’s unable to hit that target, it aims to at least reduce IT helpdesk calls by 80%. Part of the strategy to get there is the use of automation, predictive analytics and virtual assistants. “The traditional service desk is always reactive,” says Zakir Mohammed, manager of artificial intelligence and automation at Toyota. 

Looking at the experience employees can go through on a traditional IT helpdesk, he says it can take days, or even weeks, for support personnel to respond and fix an IT problem. For instance, if someone needed a new piece of software, they would need to raise a helpdesk ticket. In Mohammed’s experience, IT support would contact the individual who raised the ticket two to three days later, then block out a slot of 30 minutes or an hour to install the application.

“There was a lot of reactiveness happening,” he says. “At some point, our employees gave up opening tickets and started suffering in silence. We decided the traditional way of running a helpdesk was not sustainable. We needed to have a proactive approach.”

The company is a Gartner client, and Gartner’s digital employee experience market research recommends tools that conform to industry standards. Gartner defines DEX management tools as software that measures and continuously improves the performance of employee sentiment towards company-provided technology.

The tools tend to offer near-real-time processing of aggregated data from endpoints, applications, employee sentiment and actionable insights, which, according to Gartner, can power self-healing automation and enhance employee interactions with self-service portals and chatbots. Gartner says DEX tools also help IT support, asset management, procurement and other teams whose work depends on reliable information.

“Some of the tools and technology we evaluated looked very promising,” says Mohammed. “But the tool we currently use is Nexthink, which aligns with our requirements.”

Instead of waiting for IT issues to be manually triaged, Toyota’s 100,000 staff members now benefit from the IT department using Nexthink’s DEX technology to proactively detect, diagnose and remedy IT issues across endpoints.

Getting started with DEX

Toyota initially began a small-scale pilot of Nexthink with 100 users. Metrics were collected, enabling the IT team to understand the issues the pilot users were experiencing.

One of the challenges Toyota faced was that while it had deployed advanced observability tools to monitor business applications and IT infrastructure such as storage, the company lacked the tools required to monitor users, the performance of their devices, their experience of the IT they used and their overall sentiment.

Given people are considered a business’s most important asset, Mohammed believes it’s important to measure their experience of the IT they require to do their jobs.

Having evaluated a sample of 100 employees, he says: “What we saw was eye-opening. There were so many issues.”

Toyota then scaled up the proof of concept to 30,000 users. This step involved using automation. “We deployed to 30,000 users,” says Mohammed. “We were not only collecting the information, but we also started automating.”

Discussing the benefits of the roll-out, he says Toyota now has visibility of user devices, which helps the company offer a seamless digital experience and automates certain helpdesk tasks. 

Nexthink is also being used for predictive maintenance, such as replacing laptop batteries before they die. “If the performance of a certain type of battery is going from 80% to 60% in the next six months, these batteries may require replacement,” says Mohammed. “This is great information for the IT delivery team. It means they not only buy the batteries in advance, but can also proactively replace them before the old battery dies.”

Another way Nexthink is being used is in software reclamation. “There are tonnes of software sitting on laptops and we’re paying software licences for them,” he says. “Nexthink is able to check if the software has been used in the past 90 days. This information can then be used to send an automated message to ask if the application is still required. One click and it’s automatically reclaimed by the IT software library.”

The final piece of the DEX story at Toyota is the use of a virtual assistant. “We want to make it like a ChatGPT for Toyota, so that employees can submit a request and it does the work behind the scenes,” says Mohammed.

In effect, the virtual assistant is used to parse free text entered by users and translate these requests into actions that can be sent to Nexthink.

Another use of the virtual assistant is to enable users to request software directly. “If you need PowerBI, it connects behind the scenes with Nexthink, picks up the software and installs it,” he says. “You don’t have to do anything. Once the install is done, you get a notification saying that your software is ready.”

If Toyota’s goal is to reduce IT helpdesk calls, the ability for a user to have IT problems proactively resolved via a virtual assistant, or perhaps use it to request new software, shows where the digital employee experience is heading. 

Source

Posted on by

Microsoft 365 price hike: Are you forced to pay more even if you don’t want AI?

If you thought Microsoft’s forcing Windows 10 users to upgrade to Windows 11 to keep using Microsoft 365 Office apps was bad, you haven’t seen anything yet. Microsoft will further annoy many of its customers by embedding Copilot AI into all Office apps and charging an extra $3 per month for it. Microsoft announced the price hike for the Office 365 apps in a blog post explaining the changes.

Yes, AI development is expensive, and I absolutely agree that we, as end users, have to pay for access. That’s why I’m a ChatGPT Plus subscriber and won’t ditch that $20/month subscription anytime soon. If anything, I’m ready to pay for additional AI products that might improve aspects of my life. Take Apple Intelligence; I’ll keep buying expensive iPhones, iPads, and Macs, which will pay for Apple’s AI.

However, as a Microsoft 365 subscriber who has no interest in Copilot AI at this point, I’m not too thrilled about potentially having to shell out an extra $3 per month for my Family subscription. AI should be optional rather than mandatory in all apps. So, is there a way to keep your current subscription price if you’re like me and you don’t want Copilot AI? Well, it’s complicated.

Microsoft is forcing Copilot AI on millions of users

Microsoft 365 Personal and Family subscribers will get Copilot AI and a new Designer AI image generation app in most markets. This will lead to a price hike of $3 per month in the US, Microsoft’s first price hike for the productivity bundle subscription in 12 years. You’ll have to check your local Microsoft 365 portal to see the price increase in your local market.

Tech. Entertainment. Science. Your inbox.

Sign up for the most interesting tech & entertainment news out there.

By signing up, I agree to the Terms of Use and have reviewed the Privacy Notice.

Microsoft says that most of its 84 million consumer subscribers will have access to Copilot in Office apps whether they want it or not.

The new Copilot AI integration is separate from the Microsoft Copilot Pro subscription, which costs $20 per month. Copilot will be available in Word, Excel, PowerPoint, Outlook, OneNote, and the new Designer app.

To use the latter, Microsoft 365 Personal and Family subscribers will use AI credits that come with their subscription. The monthly allotment of AI credits applies to all Office apps and should be “enough for most subscribers.”

Let’s appreciate the good things

Microsoft knows that Copilot shouldn’t always be active in Office apps, so it’s giving users the ability to turn it off. That’s good news for students who rely on Office apps for taking notes during class.

Also important here is Microsoft’s commitment to privacy. I might not like the price hike because I don’t plan to use Copilot AI in Office apps anytime soon, but I appreciate the fact that Microsoft will not use any Office app AI data to train its models:

To protect your privacy, we do not use your prompts, responses, or file content (such as Word documents or Excel spreadsheets) when you use Copilot in the Microsoft 365 apps to train our foundation models.

You can opt out, for now

Microsoft is aware that not all Office app users will want access to Copilot AI, so there are ways to opt out and keep your current subscription price. That’s another thing I can appreciate. But there are big caveats here.

First, you have to be an existing Microsoft 365 subscriber to opt out of the Copilot AI price hike. Second, you must enable recurring billing to avoid the price increase. Those who don’t have it enabled, such as myself, won’t be able to stick with the non-AI versions of the apps.

Users with recurring billing can switch to a Basic plan, or they can keep their current plan as it exists today by switching to the new Personal Classic or Family Classic plans “for a limited time.” Once these plan options disappear, you’ll only have access to Copilot AI plans.

In other words, you’ll still be forced to pay for AI you might not want if you miss the opportunity to grab one of these limited-time plans.

Finally, there’s another big issue with the non-AI plans here. Microsoft says it will maintain the Basic and Classic plans “as they exist today,” but you risk not getting new features. For “certain new innovations and features you’ll need a Microsoft 365 Personal and Family subscription,” Microsoft says. Therefore, you’ll need to pay that extra $3 per month, or whatever it converts to in your local currency.

Should you cancel Microsoft 365?

Don’t get me wrong, I don’t think access to Copilot AI is a bad thing. It’s certainly a great tool and much more useful than, say, AI chatbots ruining WhatsApp. As you can see in the examples above, Copilot can be quite helpful in all sorts of instances using Office apps.

However, it should be up to the customer to choose to use AI. There’s no reason for Microsoft not to continue supporting non-AI Office apps in the future other than greed. That’s what it looks like, at least.

Also, since I’ve defended Netflix price hikes in the past, I’d do the same with Microsoft 365 prices if they were to go up.

My first reaction wasn’t to cancel my subscription or opt for the non-Copilot Office experience. I wanted to ask family members in the group if they wanted any built-in AI access. But I can’t even do that, considering that Copilot AI will only be available to me, the subscription owner.

Source

Posted on by

Meta is about to ruin WhatsApp with AI bots no one wants

Of all the generative AI assistants out there, Meta AI must be the most annoying for the simple fact that Meta is shoving it down our throats. No app is safe, from Facebook to Messenger to WhatsApp to Instagram. Meta AI is there whether you want it or not, and there’s no way to deactivate it.

Meanwhile, ChatGPT is entirely optional, not that OpenAI can really force it on anyone. Apple’s Apple Intelligence is also optional; you don’t have to use it even if you have access to it. Then there’s Google Gemini, which is baked into many Google products but doesn’t feel as intrusive as Meta AI. The same goes for Microsoft’s Copilot.

The worst part about Meta AI is that Meta isn’t done ruining its apps with overdoing the AI presence. We’ve just learned of AI profiles coming to Facebook and Instagram, which is extremely annoying. It gets worse; Meta will now give AI bots prime plans inside WhatsApp, a feature that nobody really asked for from the one Meta app that’s actually useful.

WhatsApp is the world’s largest chat app. It works on iPhone and Android and supports end-to-end encryption across platforms. That’s the only reason I’m still using it. That, and the fact that Meta relented on its annoying WhatsApp policy change a few years ago.

Tech. Entertainment. Science. Your inbox.

Sign up for the most interesting tech & entertainment news out there.

By signing up, I agree to the Terms of Use and have reviewed the Privacy Notice.

Come to think of it, the only reason WhatsApp is so good and still encrypted, is that the app was built this way well before Meta bought it for a small fortune.

The last thing I want in WhatsApp is support for AI chatbots. Yes, it’s great that you can save a ChatGPT number to chat with the chatbot from WhatsApp, but that’s optional.

Say that Meta feels like it has to put AI bots in WhatsApp to expose more people to Meta AI and potentially make some money in the process. I still hate the idea of WhatsApp getting a dedicated AI menu. That’s wasted screen real estate right there. It’s a feature I’ll never use, and I’m sure others will be equally uninterested.

The new AI bots menu in a WhatsApp beta release for Android.The new AI bots menu in a WhatsApp beta release for Android. Image source: WABetaInfo

Meta is testing the new interface in an Android beta version of WhatsApp. Always reliable WABetaInfo surfaced the image above that shows the new AI tab replacing the Communities tab. That menu, which might actually be useful, is merging with the Chats tab.

The new AI tab will include all sorts of AI chatbots to talk to, including third-party models that can talk to you about specific topics.

I don’t doubt that some WhatsApp users will want to use these services. I say that as a longtime ChatGPT user who chats with OpenAI’s chatbot about all sorts of things daily. But I absolutely hate the idea of any AI product being forced on me the way Meta is doing with Meta AI.

WhatsApp is especially important to me as I use it to talk to many people. It’s not just Android users in my family or friends group that like WhatsApp; plenty of iPhone owners prefer the platform over iMessage. AI isn’t needed. Or if it is, it should be hidden somewhere and accessible on demand.

It might get even worse than that. WABetaInfo found evidence in a different WhatsApp beta version that Meta wants to let users create custom AI chatbots right inside the app. The process might be similar to what’s already available on Instagram.

Support for custom AI bot creation in a WhatsApp beta release for Android.Support for custom AI bot creation in a WhatsApp beta release for Android. Image source: WABetaInfo

The feature resembles the custom chatbots available in ChatGPT and Gemini, so it’s not entirely surprising. But, again, it’s not something I want to clutter a key app like WhatsApp.

I don’t see any value in adding AI bots to WhatsApp or supporting the creation of custom ones.

Remember that if left unchecked, some custom AI chatbots might be harmful, especially when certain types of users are exposed to them. And it’s not like Meta is improving its content moderation policies, so we have no idea how it’ll police this universe of AIs it’s bringing to apps like WhatsApp and Instagram.

I can only hope that Meta will not bring these features out of beta, but that’s just wishful thinking. If anything, I take some solace in knowing that it’ll take longer for Meta to deploy the AI changes to WhatsApp in Europe.

Source

Posted on by

Samsung just hallucinated that it will become the global AI leader in 2025

Samsung was the first big smartphone vendor to launch a flagship phone with AI at the core of its marketing efforts. Last year’s Galaxy S24 series introduced the Galaxy AI suite of features. Samsung followed with the Galaxy Z Fold 6 and Flip 6, which got additional AI capabilities. Samsung then extended Galaxy AI support to older flagship devices. And in a few weeks, Samsung will launch the Galaxy S25 series, which should introduce even more Galaxy AI novelties.

But Samsung leadership is hallucinating worse than an AI program ever could about Samsung’s global role in genAI. In a New Year’s address, Samsung Electronics CEO and Vice Chairman Han Jong-hee and DS Division Vice Chairman Jeon Young-hyun addressed Galaxy AI, saying that Samsung should become the undisputed leader of device AI this year.

“Now is the time for bold innovation that goes beyond the existing success methods as we face an inflection point in AI technology,” the execs said, according to a machine-translated Samsung release. “Let’s establish ourselves as a clear device AI leader this year through advanced intelligence.”

The goal of becoming the undisputed AI leader is noble. It’s what you’d expect key execs to say ahead of a busy year when AI will continue to dominate the tech world. It’s also something officials at other leading tech companies could say, considering AI is the main priority right now.

Tech. Entertainment. Science. Your inbox.

Sign up for the most interesting tech & entertainment news out there.

By signing up, I agree to the Terms of Use and have reviewed the Privacy Notice.

But Samsung is nowhere close to being a leader in AI, and I don’t see it happening in 2025 either. The main problem with Samsung’s Galaxy AI approach is that it doesn’t have a meaningful model of its own to power the genAI tech on phones like the Galaxy S24 and S25.

Using Google's Circle to Search AI feature on the Galaxy S24 Ultra.Using Google’s Circle to Search AI feature on the Galaxy S24 Ultra. Image source: Samsung

Galaxy AI is a mix of AI technologies. Google’s Circle to Search is a good example. Also, Galaxy S25 phones are rumored to come with free Google Gemini Advanced, Google’s best version of Gemini AI.

I’ll also point out that Samsung’s upcoming XR devices, Project Moohan and unnamed AR smart glasses, will work on Google’s Android XR platform, with Gemini playing a key role. I expect Galaxy AI to be part of the picture for both types of products because Samsung can’t AI on its own.

Samsung doesn’t have an alternative to ChatGPT or Gemini. If it is working on Bixby upgrades and Gauss upgrades, matching these AI models will take a long time.

Also, Samsung doesn’t have a desktop presence. ChatGPT is my primary AI tool right now, and I use it across devices. Most of the time, I access it on my Mac rather than a mobile phone.

OpenAI and Google have better models. Meta, Claude, and Microsoft also have AI tools that are more advanced than Samsung’s. Apple is working on a Siri LLM that will behave like ChatGPT and has incorporated ChatGPT into Apple Intelligence on the iPhone.

As for on-device AI, Samsung might have been the first to push AI on mobile devices with Galaxy AI, but it’s not the only one. Google is doing it with Pixel phones and Android in general. Apple laid out an even better vision of on-device AI with Apple Intelligence this year, which Samsung doesn’t appear to be able to match.

Samsung's Project Moohan Android XR headset.Samsung’s Project Moohan Android XR headset. Image source: Samsung

Apple Intelligence might be behind Galaxy AI and other rivals, but Apple has something rivals can’t match: a massive base of devices that can use Apple Intelligence, and the list is growing rapidly. Once Apple Intelligence matures, Apple could very well become the undisputed device AI leader.

Speaking of Apple’s AI vision, Samsung has yet to match what Apple wants to do with iPhones. It’s not just about text and notification summaries, text generation, wallpaper generation, photo editing, and translation. It’s about Siri becoming a more useful assistant by accessing on-device contextual information about the user.

Apple has a plan, at least; one that Samsung might follow. Samsung’s Galaxy AI teasers during the Fold 6 and Flip 6 launch event revealed the company is working on a similar vision. But Samsung waited for Apple’s Apple Intelligence reveal before it unveiled its own plans.

I’ll also point out that Apple Intelligence is designed to offer more on-device AI features and better privacy for cloud-based AI than Galaxy AI can. Turn off Galaxy AI on your phone right now, and you’ll lose many of its useful features. Samsung has yet to match Apple’s Private Cloud Compute, a private cloud-based AI system.

What I’m getting at is that it’ll take years for any company to become the undisputed leader in device AI. If that ever happens. And it’s way too early for Samsung to call for that title, especially considering its massive reliance on partners like Google.

Also, suppose the Samsung execs only want the company to sell as many products that can run third-party AI programs within Galaxy AI. In that case, that still doesn’t qualify as being the undisputed leader of device AI.

Source

Posted on by

iOS 19 will reportedly work on every iPhone that supports iOS 18

Reports in the past few weeks claimed that Apple’s main focus remains Apple Intelligence. The company was working on the now-released iOS 18.2, which brings ChatGPT integration to the iPhone and iPad, and iOS 18.4, which will deliver a smarter Siri.

Those reports said work on iOS 19 had been delayed. Therefore, some iOS 19 Apple Intelligence features might see delays similar to what happened this year.

As for the non-AI features in iOS 19, I said at the time that I expect Apple to ship novelties in next year’s operating system. After all, Apple Intelligence will only work on the iPhone 17, iPhone 16, and the iPhone 15 Pros.

While we’re yet to find out the big non-AI features of iOS 19, there is good news for iPhone owners that use older models. A leak says that all the models that can run iOS 18 will also run iOS 19 next year. The only compatibility change will impact the iPad, as certain models will lose support for iOS 19.

Tech. Entertainment. Science. Your inbox.

Sign up for the most interesting tech & entertainment news out there.

By signing up, I agree to the Terms of Use and have reviewed the Privacy Notice.

French blog iPhoneSoft learned from a purported source inside Apple that development work on iOS 19 had not started for this particular person. That’s a first for the leaker compared to previous years. Instead, they’re supposedly working on iOS 18.x updates and visionOS.

This person reportedly learned the iPhone models that will run iOS 19 next year, telling the blog that all the iPhones compatible with iOS 18 will also support the next OS upgrade. iPhoneSoft listed all the iPhones that will support iOS 19 next year, including the unreleased iPhone 17 and iPhone SE 4 models:

  • iPhone 17, 17 Air, 17 Pro, 17 Pro Max (2025)
  • iPhone 16, 16 Plus, 16 Pro, 16 Pro Max (2024)
  • iPhone 15, 15 Plus, 15 Pro, 15 Pro Max (2023)
  • iPhone 14, 14 Plus, 14 Pro, 14 Pro Max (2022)
  • iPhone 13, 13 Mini, 13 Pro, 13 Pro Max (2021)
  • iPhone 12, 12 Mini, 12 Pro, 12 Pro Max (2020)
  • iPhone 11, 11 Pro, 11 Pro Max (2019)
  • iPhone XS / XS Max (2018)
  • iPhone XR (2018)
  • iPhone SE (4th generation) (2025)
  • iPhone SE (3rd generation) (2022)
  • iPhone SE (2nd generation) (2020)

Regarding iPadOS 19, Apple will stop models for some older iPads that can still run iPadOS 18. You’ll need an iPad with an A12 chip or later to run iPadOS 19 next year. Here’s the list of supported iPads: 

  • iPad mini (5th generation or later)
  • iPad (8th generation or later)
  • iPad Air (3rd generation or later)
  • iPad Pro (2018 or later)

While these early iOS 19 and iPadOS 19 compatibility claims make sense, there’s no way to confirm any of them. Apple will hold its WWDC 2025 event next June. We’ll see plenty of iOS rumors by then, which will shed further light on the features Apple might be working with and the devices that will support them.

Source

Posted on by

A jobseeker’s guide to using AI and what it means for employers

Artificial intelligence (AI) is a powerful tool to help jobseekers find roles and make their applications, with ever more people using it. Multiple published surveys have suggested this figure could be as high as 50% of applicants. But while AI is undoubtedly a great support tool, it can create issues if individuals use it to present a misleading impression of themselves and their capabilities. So how can it best be used – and what are the do’s and don’ts for jobseekers to think about?

At the same time, the growing use of AI presents new challenges for employers. In some cases, it is dramatically increasing the number of applications for employers to work through. Figures from the Institute of Student Employers show a 59% rise in the average number of applications being received for graduate jobs (140 per position) with recruiters in higher-paid and growth sectors, such as digital and IT, receiving as many as 205 applications per vacancy – and at Harvey Nash, we are seeing as many as 500 in some instances. The institute says AI is the driver of these increases. Moreover, can employers trust that applications actually represent candidates faithfully and honestly? In this article, I’ll highlight some advice points for them to consider too.

How jobseekers can use generative AI

There are multiple ways in which AI tools can help jobseekers in their efforts to land that dream role. Some of the best-known tools include ChatGPT, Microsoft Copilot, Gemini and Bard – with many other more specialised tools available for job searches and application support.

Tailored applications

Understanding descriptions: Generative AI tools can instantly summarise complex job descriptions, helping candidates quickly understand core responsibilities and requirements, allowing them to tailor their applications effectively.
Highlighting relevant experience: By extracting key information from job descriptions, candidates can emphasise relevant skills and experience in their CVs and cover letters.

CV improvement and optimisation

AI-driven CV refinement: Jobseekers can use generative AI to enhance their CVs. Tools can suggest improvements, optimise formatting and ensure that critical details stand out.
Keyword optimisation: AI can identify relevant keywords for specific roles, improving a CV’s chance of passing automated screening tools.

Interview preparation

Mock interview simulators: AI-powered simulators can help candidates better prepare for interviews. By posing common interview questions and provide feedback, they help to build a candidate’s confidence and enhance their overall interview performance.

Job matching

AI-powered job search: Many tools can match candidates with suitable roles based on their skills and experience. This streamlines the job search process and helps candidates identify the roles they are most suited for.

The benefits that AI can bring

Using AI tools in this way brings a number of benefits to jobseekers, most notably:

Efficiency: Generative AI accelerates tasks like summarising job descriptions, refining CVs and preparing for interviews.
Productivity boost: AI can act as a ‘work buddy’, helping candidates better manage and prepare when applying for multiple vacancies.
Improved quality: AI can help candidates better communicate their strengths and present themselves more effectively, increasing their chances of being shortlisted or interviewed.
Advanced options: Many AI tools are freely available, but there are also paid-for versions of tools that offer even greater functionality and have a greater ability to learn from previously produced content to reflect an individual’s tone of voice or style of language.

Do’s and Don’ts for candidates

While these are all compelling benefits, nevertheless the use of AI does present various potential issues. AI tools can have the effect of making everyone’s applications and the way they present information look the same. There is a danger of losing individuality as applications become more vanilla and standardised. Here are some advice points accordingly:

Do…

  • Use your own words and language as much as possible to keep it authentic and bring out your own character. If using AI to create your CV, stand back from it and ask yourself whether the structure of it is bringing out your unique qualities and experience effectively.
  • Avoid generic phrasing that feels stilted or impersonal – otherwise there is a danger of a “sea of sameness”.
  • Answer interview questions/tasks on your own. You may want to use AI to refine them afterwards, but always start with your own answers. It’s your own knowledge and ability that you’re being assessed on – and you might get caught out later on.
  • Use AI as a support tool – not to do the whole job for you. It can help you make the process quicker and more efficient, but shouldn’t become a substitute for you putting in the appropriate level of effort yourself.

Don’t…

  • Lie or exaggerate to give a false impression, otherwise there is a danger of AI becoming like ‘catfishing’ for job applications. Checks in the process later on will almost certainly expose any untruths.
  • Use AI to send off reams of untargeted applications on the off-chance you might be successful. This will ultimately waste your own time as well as the employer’s.
  • Use Americanisms and American spellings (if you’re in the UK) which many generative AI tools are programmed with. Adapt what AI produces so that it’s suitable for the market you are in.
  • Pass off AI-generated answers or content as your own. You need to build relationships with recruitment agencies and prospective employers and will lose their trust if they realise you have been leaning excessively on AI.

What does this mean for employers and recruiters?

The use of AI by candidates and jobseekers is something that employers have become increasingly aware of. There is no problem in principle with a candidate using AI – indeed, it shows initiative and with many organisations embedding AI into their own process and systems, it would likely often be seen as a positive. Nevertheless, it is having some impacts that employers need to manage.

Firstly, as I have noted, AI is ramping up the number of applications that employers are receiving, almost becoming a barrage in some instances. This creates a workload issue, with teams having to sift through many more applications, cover letters and CVs to produce their shortlist of candidates.

Secondly – and more seriously – AI is making it harder for employers to really know how capable a candidate is, given that applicants may use AI to smarten their CVs, word their covering letters, answer questions on application forms, and assist them with remote/take-home tests and technical exercises like coding challenges.

There are several ways that employers can manage the situation, in particular:

Review your assessment techniques: Look across the questions and tests you set candidates, and consider whether you should introduce more open-ended questions that are harder for AI tools to answer authentically. Use real-world scenarios and situational questions that require human experience to respond to. Also think about using more on the spot tests that candidates take in your offices or assessment centre rather than remotely.
Upskill your teams: Think about providing training for your in-house recruitment team and hiring managers to understand how AI is changing the landscape, and what to look out for. This training could include interview techniques – how to effectively probe candidates on information they have given or skills/experience they say they have.
Consider the recruitment agency option: Depending on the number of vacancies your business has and the number of applications you receive, a good recruitment agency could be a significant support. Experienced recruiters can take the burden away from already stretched in-house teams. Recruiters should be well-versed in the phenomenon of AI and have the tools to screen and assess applications, CVs and other materials. They should also speak or communicate directly with candidates of potential interest (face-to-face, on the phone or video call, and/or via email) before putting them forward for interview – making sure that they are who they say they are and have the skills and capabilities to match.

It is fair to say that AI presents the biggest challenges for enterprises running large-scale recruitment activities such as graduate schemes or other high-volume intakes. These are more prone to candidates trying to ‘game’ the system supported by AI. But it is presenting issues for all employers to be aware of.

For all these challenges, there are nevertheless several benefits that AI can bring employers too. AI tools can help prepare candidate information packs (and agency briefs) more easily and quickly. They can score various types of tests automatically. And AI can be used to support the diversity and inclusion agenda – scanning draft job adverts and role descriptions to identify whether they are optimally worded, including considering the needs of specific groups such as people with disabilities or those who are neuro-diverse.

Making AI beneficial to all

Used well, AI can significantly help both sides – jobseekers and employers alike. One thing is certain: it is here to stay, and indeed can be expected to dramatically grow as tools become more widely available and functionality continues to mature. 

This just underlines the importance for both individuals and employers to understand the dynamics at play and observe the emerging etiquettes – in order to create benefits for everyone while minimising the threat of downsides.

Emma Gardiner is regional director for UK North at Harvey Nash

Source