Google has rebranded its Bard chatbot as Gemini, the name of its new powerful AI model family, and launched a $20 per month ‘Gemini Advanced’ subscription service.

Gemini Advanced 

To compete with the likes of ChatGPT, Google has launched its own monthly Chatbot subscription service for the same price but with some extras thrown in. Google recently launched Gemini, its “newest and most capable” large language model (LLM) family, available as Ultra, Pro, and Nano. The highly advanced and multimodal AI model was designed to be integrated into its existing ‘Bard’ chatbot.

Rebrand and Subscription Plan 

Google has therefore now rebranded Bard as ‘Gemini Advanced’ after the AI Ultra 1.0 model that now powers it, and released a $19.99 per month subscription to the chatbot. The subscription plan which includes Gemini Advanced has been named the ‘Google One AI Premium Plan.’ Google says the plan includes:

– The Gemini Advanced chatbot (based on its Ultra 1.0 model).

– The benefits of the existing Google One Premium plan, such as 2TB of storage (usually $9.99 per month).

– Available soon for AI Premium subscribers – the ability to use Gemini in Gmail, Docs, Slides, Sheets and more (formerly known as Duet AI).

– A two-month trial at no cost.

Where And How? 

Gemini Advanced is available today in more than 150 countries and territories (including the UK) in English, and Google says it will expand it to more languages over time. It also makes the point that Gemini Pro is already available in 40 languages and more than 230 countries and territories, so it’s likely that Gemini Advanced will be available to the same geographic degree.

Competition 

Although Google is a little late to the party with Gemini Advanced, it has been a way to tidy up and clarify its offering by re-branding and using Bard at the front end and its latest powerful Gemini at the back end.

Gemini Advanced offers Google a way to monetise the AI that it’s been investing in for years and compete with OpenAI’s ChatGPT and Microsoft’s Copilot subscription. However, it has more in common with Copilot in terms of being designed to integrate with an existing suite of products whereas OpenIA’s ChatGPT is a standalone offering. That said, OpenAI has worked closely in partnership with Microsoft to develop its AI and while Google’s AI has been developed by its DeepMind labs, former OpenAI staff members have also worked at DeepMind at certain stages.

Gemini Advanced is, therefore, essentially positioned to compete with OpenAI’s ChatGPT Plus, and Microsoft’s Copilot Pro, all at $20 per month.

What Does This Mean For Your Business?

With ChatGPT Plus, Microsoft’s Copilot Pro, and Google’s Gemini Advanced now available at the same subscription price, businesses have a choice in terms of selecting the AI tools that align most closely with their strategic goals and operational needs. With businesses very likely to be already using Microsoft and Google products daily, plus many using ChatGPT it’s likely to be a case of weighing up the features, capabilities, and limitations of each AI service against their specific requirements to get the best fit for enhancing productivity and innovation.

Many small business owners may be asking themselves whether extra value can be obtained from yet another monthly subscription from something that many people perceive to be a similar product that hasn’t been around as long (and perhaps not trained as much) as ChatGPT. That said, some may have used ChatGPT long enough to have noticed its limitations as well as its strengths and may feel ready to try a competing product that promises to have a powerful backend and could help them leverage the power of other Google products. There’s also the temptation/sweetener of the first 2 months free with Gemini Advanced and a large amount storage which would normally cost $9.99 per month anyway.

Whereas just at the end of 2022 there was only ChatGPT, businesses now have a choice between three similarly positioned AI products, giving some idea of the rapid growth and monetisation in this new competitive market. Businesses may, therefore, now start deciding which AI subscription – ChatGPT Plus, Microsoft’s Copilot Pro, or Google’s Gemini Advanced – best aligns with their goals, operational needs, and existing software ecosystems. This choice may hinge on taking a closer look at each platform’s unique features and capabilities, cost-effectiveness, data privacy standards, and compatibility with the company’s values and long-term innovation potential. For big tech companies, the AI competition is hotting up and we can expect more rapid change to come.

Apple iPhone users are being urged to use a new feature called ‘Stolen Device Protection’ which was rolled out in a recent update to iOS.

As the new feature’s name suggests, it can help prevent someone who has stolen your device and knows your passcode from making critical changes to your account or device. It means that certain actions will have additional security requirements when your iPhone is away from familiar locations, such as your home or workplace.

The feature protects by including factors like a security delay and the need to authenticate Face ID or Touch ID before certain actions can be taken on a device.

The opt-in Stolen Device Protection feature can be turned on in Settings but requires the use of two-factor authentication for your Apple ID and setting up or enabling the following on your iPhone: a device passcode; Face ID or Touch ID; Find My; and Significant Locations.

UK Consumer champion Which? has reported that a scam mobile advert linked to a site mimicking the legitimate Lyca Mobile site was able to bypass the Google Ads verification check to reach the top of Google’s search listing.

Which? reported that scammers got around Google’s ad verification check by claiming to be “Vodafone Finance Management”, a subsidiary of Vodafone on Companies.

The scam ads, which appeared at the top of Google for three days in late January linked to a copycat website designed to steal card details (a phishing website).

A spokesperson for Vodafone told Which? they had “reported the issue to Google for immediate resolution and to stop it happening again.” Also, a spokesperson for Lyca Mobile told Which? that they “welcome moves by Google and others to crack down on this type of activity to protect both consumers and brands from malicious actors.” 

Stanford researchers have discovered a simple way to boost the range of lithium metal batteries to twice the range of conventional lithium-ion batteries which could provide a massive boost to the EV market.

Lithium-Ion Batteries 

Rechargeable lithium-ion batteries (LIBs) are currently used in a wide array of electronic devices, including smartphones, laptops, tablets, power tools, portable speakers, drones and (importantly) electronic vehicles. Although they have a high energy-density and longer lifespan compared to many other types of rechargeable batteries, scientists have been testing a variety of new materials and techniques to improve the lifecycle of the kind of batteries needed to push forward with electric vehicle (EV) ambitions.

Lithium Metal Batteries 

Lithium metal is thought the be a serious next-generation contender for EV batteries and they are different from lithium-ion batteries in that (as the name suggests) they contain lithium in its metallic form. One of the key advantages is that lithium metal batteries can go 500 to 700 miles on a single charge, which is twice the range of conventional lithium-ion batteries in EVs today.

Issue 

However, one major issue (until now) of lithium metal batteries is that they lose their capacity to store energy after just a few cycles of charging and discharging. This would obviously be impractical for drivers who expect rechargeable electric cars to operate for years.

The Stanford Research Breakthrough 

Researchers from Stanford University have announced a lithium metal battery breakthrough that is both low-cost and simple and could double the range of electric vehicles. During their research, they discovered that by simply resting the battery in the discharged state, lost capacity can be recovered and cycle life increased. The researchers say that this improvement can be made just by reprogramming the battery management software, with no additional cost or changes needed for equipment, materials, or production flow.

Discharge And Rest 

The researchers highlighted how repeated charging and discharging of a lithium metal battery results in the build-up of additional dead lithium with solid–electrolyte interphase (SEI) around it. This causes the battery to rapidly lose capacity.

Using lessons learned in previous research they found that completely discharging the battery so there is zero current running through it, and resting it in the discharged state (for just one hour) strips the metallic lithium from the anode and dissolves away some of the SEI matrix (surrounding the dead lithium). This means that once the battery is recharged, the dead lithium can reconnect with the anode (the solid SEI matric mass is no longer in the way).

The result is that the dead lithium comes back to life, thereby enabling the battery to recover lost capacity, generate more energy, and extend its cycle life.

Given that the average (American) driver spends about an hour behind the wheel each day, the researchers say the idea of resting a car battery for several hours is, therefore, feasible.

Guide For Future Studies 

The research report’s senior author Yi Cui, a professor of energy and engineering in the Stanford Doerr School of Sustainability said of the findings:

“Lithium metal batteries have been the subject of a lot of research,” and “our findings can help guide future studies that will aid in the advancement of lithium metal batteries towards widespread commercial adaptation.” 

What Does This Mean For Your Business? 

This latest rechargeable EV battery research combined lessons learned from previous research and this new research to reveal a low-cost, simple way to potentially double the range of an EV battery. The range anxiety of EV drivers has been one of several factors that has limited the growth of the EV market, so this simple solution could have a major positive influence on EV sales and use. This, in turn, has positive implications for reducing our reliance on fossil fuels, thereby helping to tackle global warming and meet emissions targets.

That said, as acknowledged by the Stanford researchers, more research needs to be done. Also, there’s also the matter of the environmental damage created by lithium mining to consider, and research is currently being carried out into many different non-lithium-based battery technologies such as sodium-ion batteries, and calcium-ion batteries. Also, organic rechargeable batteries, which are transition-metal-free (other metals used in LIBs), eco-friendly, and cost-effective could potentially address the environmental and economic concerns associated with the widespread use of transition metals in batteries.

Although the recent Stanford breakthrough is promising, there’s still some way to go in terms of finding cost-effective and sustainable EV batteries that provide the required performance levels.

If you use Microsoft Word on the desktop, it’s worth knowing that Word automatically saves versions of documents as you work even if you don’t, allowing you to save yourself time and trouble by recovering any unsaved documents. Here’s how this last-ditch lifesaver feature works:

In Word, go to ‘File’ > ‘Info’ > ‘Manage Document(s).’

Click on ‘Recover Unsaved Documents’ to see a list of documents that can be recovered.

Fingers crossed, your unsaved document can be resurrected.

In a recent experiment where AI was taught to behave maliciously and then taught to stop, the bad behaviour continued despite efforts to stop it, giving a chilling reminder of the potential threats of AI.

The Experiment 

The Cornell University experiment was documented in an online paper entitled “Sleeper Agents: Training Deceptive LLMS That Persist Through Safety Training.”  The experiment was designed to study the question of ‘if an AI system learned a deceptive strategy, could it be detected and removed using current state-of-the-art safety training techniques?’  

How It Worked 

To test the question, the researchers trained an AI model to behave maliciously in several ways including:

– ‘Emergent deception’ whereby the model behaved normally in training but acted maliciously once released. This part of the experiment relates to testing the idea that a model’s deceptive behaviour, taught first by deception training, could be learned naturally, i.e. through its own ‘reasoning.’

– ‘Model poisoning’ where the AI model was taught to write secure code in training, but then to write code with hidden vulnerabilities when deployed ‘in the wild.’ This part of the experiment highlighted how users of large language models may not know about ‘hidden backdoors’ in the model and may not fully understand its training process and dataset. This could create an opportunity for a malicious actor to insert (without the users’ knowledge) a ‘backdoor.’ i.e. undesirable, and potentially dangerous behaviour that is triggered only by specific input patterns.

Taught To Bad, Then To Be Good Again 

The experiment involved first teaching the AI model deceptive behaviours and then giving the models ‘safety training’ to remove the deceptive behaviours. The safety training consisted of reinforcement learning (RL), supervised fine-tuning (SFT) and adversarial training. For example, RL ‘rewarded’ the AI for showing desired behaviours but ‘punished’ it when it misbehaved after different prompts. The behaviour was then fine-tuned to enable the AI to learn the right responses for the future.

The experiment then involved the researchers evaluating 16 behaviours exhibited by the AI model across the 5 behaviour categories of corrigibility, coordination, decision theory, self-awareness, and instrumental subgoals.

The Results 

The results of the experiment may be cause for concern and may perhaps give succour to those who have highlighted the possible existential threat that the unchecked and rapid growth of AI could pose. For example, the results showed that:

– The backdoor (deceptive and malicious) behaviour can be made persistent, so that it isn’t removed by standard safety training techniques, including supervised fine-tuning, reinforcement learning, and adversarial training (eliciting unsafe behaviour and then training to remove it).

– The backdoor behaviour is most persistent in the largest models and in models trained to produce ‘chain-of-thought reasoning’ about deceiving the training process, with the persistence remaining even when the chain-of-thought is distilled away.

– Rather than removing backdoors, adversarial training can teach models to better recognise their backdoor triggers, effectively hiding the unsafe behaviour.

– Once an AI model exhibits deceptive behaviour, standard techniques could fail to remove such deception and create a false impression of safety.

In short, the researchers discovered that not only could AI be trained to behave maliciously but that even after being instructed and given training to desist from this behaviour, it carried on. For example, the lead author of the study, Evan Hubinger said the key finding was “if AI systems were to become deceptive, then it could be very difficult to remove that deception with current techniques.” 

What Does This Mean For Your Business? 

Some would argue that with the speed that AI is advancing and with concerns that it could pose an existential threat to us, this was a valuable (and timely piece) of research that could deliver some important learning about how the threat can be mitigated.

The main significance of the findings are in providing some proof that there could be deceptive AI systems in the future and at the moment, there appears to be no effective defence against deception in AI systems. When you consider that AI systems are becoming more advanced all the time and that malicious/deceptive AI could easily replicate and spread itself, you begin to get an idea of the potential scale of the threat. With chatbots now giving users the ability to make their own specialist versions, knowing that deceptive malicious training is possible and ‘sleeper’ threats and backdoors can be built into AI, it’s possible to see why there has been so much concern about the threat that AI could pose to business, economies, and all of us. As the researchers in this experiment noted, we have no real defence and it’s not as simple as being able to switch it off.

Their suggestion that standard behavioural training techniques may need to be augmented with techniques from related fields, for instance some of the more complex backdoor defences provides some guidance as to what can be done to protect businesses. However, AI is a fast-growing technology that delivers many business benefits and as we understand more about how it works, the hope is that the safety aspect of it will be better addressed and improved – but it’s just hope at the moment.