The letter, Pause Giant AI Experiments: An Open Letter, was published by the Future of Life Institute, a non-profit organization partially backed by Elon Musk. It says, “Powerful AI systems should be developed only once we are confident that their effects will be positive and their risks will be manageable.”

The signees are asking all AI research labs to halt, for at least 6 months, the training of any AI tool that is considered more powerful than the current state-of-the-art, GPT-4. They suggest using the pause to build and implement common safety protocols to minimize the potential for abuse and ensure outside experts can properly oversee ongoing research.

Notably, the letter does not advocate a full-on pause in AI-related research. It requests only a pullback from training activities that could lead to deployment of even more advanced models, the inner workings of which researchers are still struggling to understand.

TWO SIDES TO EVERY ARGUMENT

Predictably, the letter has touched off a raging debate over the ethics of AI, and our individual and collective responsibility to ensure we don’t end up living in a dystopian Hollywood nightmare.

On the one hand, it’s difficult to ignore a letter signed by names like Apple co-founder Steve Wozniak, OpenAI founder Elon Musk, and Yoshua Bengio, founder and scientific director of Mila, a Montreal-based AI research institute, 

Considering who the signees are – AI researchers and pioneers, computer scientists, academics, entrepreneurs, authors, and government officials – it would be careless to dismiss the letter as an ill-considered rant. It speaks to the litany of very real risks of allowing the technology to evolve without sufficient protections or guardrails in place.

On the other hand, it can seem somewhat disingenuous for people like Mr. Musk to spearhead a rather sudden move to freeze next-generation platform training. Mr. Musk’s Tesla vehicles, after all, use AI-based tools extensively, as does the SpaceX Crew Dragon. Many of the signees have profited handsomely from AI development – all while the planet has struggled to come up with a plan that balances development with the future safety of the human race.

The cynical among us would be forgiven for concluding that all of the signees have a significant stake in the future of AI development, and aren’t above angling for a pause to maintain their leadership and delay the entry of competing actors. 

ALL ABOARD

The sheer logistics of halting global development are also beyond the scope of imagination. Beyond having no historical precedent, there is simply no global entity with the power to ensure compliance.

And even if such a planetary AI police force existed, it would lack the ability to rein in rogue nations like Russia, China, North Korea, and Iran; all of which have invested heavily in sponsoring digital insecurity., Not all nations, after all, would agree to be part of this proposed planetary Kumbaya moment.

Indeed, if we suspend disbelief for a moment and envision this pause actually taking place, it would only serve to empower these four nations and endanger all other nations.

LISTEN TO BILL

The most cogent argument against this letter comes from none other than Microsoft co-founder Bill Gates. While he is no longer involved in its day-to-day operations, there is some irony in the fact that Microsoft has emerged as an early leader in the rush to integrate AI-based tools into its productivity products.

Microsoft has also invested heavily – some reports say more than $10 billion – in ChatGPT maker OpenAI and is providing much of the muscle to monetize the just-released GPT-4 large language model. The Redmond, Washington-based organization’s efforts are ramping up just as Google’s owner, Alphabet, pivots toward AI in an all-hands-on-deck effort – known internally as a “code red” – to inject AI technologies into its core search-based products.

Speaking with Reuters after the letter was published, Gates said a global pause simply couldn’t be pulled off, and even if it could, it wouldn’t address AI’s fundamental problems. He instead suggested focusing on the beneficial use cases for AI, while keeping an eye on what he calls “the tricky areas.”

Even before the Future of Life Institute letter was published, Gates, who has been a vocal supporter of AI, published a letter of his own, The Age of AI has Begun, in which he called for investments in AI to help address global inequity.

Gates’s comments notwithstanding, the dollars at stake underscore why any form of pause is highly unrealistic. The signees likely appreciated this even before they lent their names to the project. Their words are a profound warning to us – way more than a simple request – and the very fact that they’ve asked for a 6-month stop reinforces just how serious they perceive the threat to be. 

LEARN FROM HISTORY

We’ve seen what can happen when technology races ahead with no built-in controls or legislation. Social media provides an ideal and recent case study. For the better part of the past two decades, the rampant spread of social platforms has illustrated the costs society pays when advances in technology outpace any relevant protections.

Facebook’s Cambridge Analytica scandal, for example, rather starkly underscores what happens when profit-seeking organizations move too fast and break too many things. In this case, privacy concerns were swept under the rug within a technology and process architecture that allowed abuse to occur completely by design.

Instagram’s widely reported impact on teen mental health – and the parent company’s decision to dial up the algorithm as the research emerged – also makes the case for greater scrutiny of emerging technologies.

So temporarily slowing down AI’s rush to the future so we can put proper guardrails in place isn’t a completely bad idea. And as infeasible as it will be to implement and maintain such a global standdown, at the very least this letter should prompt a long-overdue discussion. (We wrote about this issue recently here: TikTok bans are a long overdue wakeup call for mobile security).

THE BOTTOM LINE

Business leaders are easily forgiven for wondering how to make sense of the current firestorm surrounding AI’s rapid evolution. The release of ChatGPT to the public in November 2022 has unleashed an unprecedented wave of interest – and investment. As was the case with earlier generational shifts in technology, like the PC, the web, social media, and smartphones/apps, it can be difficult for organizations to decide when and how they’ll start working with these technologies,

But as the open letters and the countering claims fly back and forth, organizations of all types and sizes, in all sectors, need to at the very least, start having their own discussions about their projected AI journeys. Even if there are no definitive answers yet, they can start asking questions about how their businesses might be impacted and how others in their competitive sector are beginning to assess this new landscape.

At the same time, they should be speaking with trusted partners to brainstorm strategies and temperature-check their plans. Ultimately, AI is software. Sophisticated software, of course, but software all the same. And since software is what we do, we’re here when you’re ready to brainstorm your own AI journey. No letter required.

The post Tech leaders pen letter demanding AI pause: are we moving too quickly? first appeared on STEP Software.

As we wrote about recently, ChatGPT’s arrival marked a significant inflection point in the tech industry, as it represented the general public’s first direct exposure to a full-on AI-based platform. Up until now, these technologies largely existed in labs.

The global release of ChatGPT turned this once-sci-fi-seeming technology into an everyday reality as millions of regular folks logged in and tried it out. ChatGPT’s public debut has understandably spurred a flurry of activity among major tech companies as they jockey for position in this chaotically accelerating space.

Much like the early days of the search engine wars a generation ago, AI-related announcements are now almost daily affairs. Here’s a quick summary of some of the major moves that have happened in the AI space since ChatGPT first showed up in our browsers:

• Microsoft ups its investment in OpenAI – Microsoft had already invested billions in ChatGPT’s creator, and after reportedly opening its wallet again to the tune of a $10 billion investment, the Redmond-based company placed a huge bet on who it felt would win this nascent war.
• Microsoft releases – then kills – Sydney chatbot for Bing – It’s one thing to release a standalone chatbot, and quite another to incorporate it into an existing platform. Microsoft’s first significant move was to incorporate ChatGPT directly into its Bing browser. The resulting chat tool, known as Sydney, helped drive a massive uptick in interest in the long-ignored search engine. But just as quickly, it became combative in conversations with a New York Times journalist – even trying to break up his marriage. 
• Google launches Bard – Perhaps no one has more to lose than Google, whose dependence on search – and search-based ad revenue – is at risk of being leapfrogged by AI-based chatbots. So, this ChatGPT competitor critically telegraphs whether the web services giant will be able to compete in this emerging domain. While the initial demo was a bit of a bust – and caused a short-lived selloff in parent company Alphabet’s stock – the company remains committed to it.
• Google announces Workspace integration – The company will start slowly, with new AI-powered features in Gmail and Google Docs that could potentially free users from the tyranny of starting with a blank screen. The AI bot, initially available only to a limited number of Workspace’s 3 billion users, will help them draft initial versions of email messages, job descriptions, and even a birthday party invitation. As availability eventually expands to a wider swath of users, Google will integrate generative AI features into additional tools, including Chat, Meet, Sheets, and Slides.
• Microsoft 365 Copilot integrates AI into Word, PowerPoint, Excel, and Teams – CEO Satya Nadella is wasting no time trying to generate a return on the company’s $10 billion investment. Business Chat, for example, generates updated status reports from data found in calendars, email messages, chats, documents, meetings, and contacts. In PowerPoint, Copilot can build comprehensive decks from a single query. In Teams, it summarizes discussion points and suggests action items – right in the middle of a meeting.
• OpenAI releases GPT-4 – The original ChatGPT was based on the GPT-3.5 large language model (LLM). LLMs are to chatbots as operating systems are to apps, and the release of the next-gen GPT-4 LLM added massive capabilities on top of the already-headline-grabbing original. Compared to GPT-3, which can only process text-based queries, GPT-4 is known as multimodal, and can process images in addition to text. For example, it can take an image of a fridge, then suggest meal plans based on the ingredients in the scene.

Thanks to ChatGPT, the entire AI space has gone supernova – almost reminiscent of the late 1990s emerging Web market. As we’ve seen with previous major generational shifts in tech, major and minor vendors alike are jockeying for position as they try to establish themselves in a market experiencing seismic change.

Make no mistake: AI is already shaking up the global tech order. As chaotic as this all seems, rushed announcements like Microsoft Sydney and botched demos like Google Bard are crucial steps as the dominant platform players try to hold onto their positions.

BUT WHAT DOES ALL THIS MEAN TO ME?

As entertaining as it can be to watch Microsoft and Google duke it out, the real battle is taking place on your laptop, tablet, and smartphone. Amid the flurry of announcements in recent months, the productivity app integration plans from both Microsoft and Google stand out as the ones that will ultimately matter most to CIOs, IT Directors, and everyday individual users.

Whatever your organization, sector, or role, the time to start working with these fast-evolving tools is now. Ignoring this revolution is simply not an option. Whether you’re a Microsoft or a Google shop – or both or neither – start evaluating the end-user productivity tools you’re currently using and assess whether or not any of these new AI-based tools are accessible.

Reach out to your people – developers, administrators, project managers, everyone who does the work that keeps your lights on – and take the time to understand their current productivity needs. Know the tools they’re using, what works for them, and where their productivity-related frustrations may be. Look for pain points – and use cases where AI can relieve their administrative burdens and free up cycles for more value-added, higher-order workflows.

Here at STEP Software, we’ve been developing smart code for decades, and we’re already having fulsome discussions with our clients about AI’s role in their future software roadmap. Give us a call if you’d like to discuss how all this AI-driven change can affect you – and how you can best prepare for it.

In the meantime, watch this blog, as we’ll continue to observe and share perspective on this exciting chapter in tech history.

The post AI chatbots go next-level – and productivity apps are their next target first appeared on STEP Software.

Artificial Intelligence (AI) has been one of those on-the-horizon technologies for what seems like forever. But public interest and engagement has been somewhat limited. That’s largely because it’s been accessible mostly to developers, researchers, and laboratories.

OpenAI’s decision to release ChatGPT to the public in November seems to have changed the game markedly. By opening up its consumer-friendly chatbot to everyday users, the organization, co-founded by Elon Musk, has sparked a global explosion of interest.

At the same time, all this newfound attention is fueling a similarly global-scale debate over the role knowledge workers will play in the future AI-driven economy – and or whether such workers will have roles, at all.

To step back, chatbots represent something of a generational leap beyond the search engines we’ve been using for the past generation. Unlike a search engine, where users punch in some search terms and in return receive a list of links, a chatbot is much more interactive.

Ask it fully formed questions, and it uses AI to formulate similarly fully formed responses. It’s like a conversation – often eerily human-like – and it can be used to generate a broad range of content – from writing business letters, emails and academic papers to song lyrics, and short stories. It can even write software code based on a prompt or figure out how to pick the lock on the front door.

So, it can do lots of things – but can it do them particularly well? A Wharton professor had ChatGPT take a final exam for an operations management MBA course, and the results may surprise you. While the chatbot did, in fact, pass the exam, the prof would have scored it between a B and a B-. Not exactly dean’s list material.

But here’s the thing: because ChatGPT is an AI-based tool, it will learn and get better over time, particularly as OpenAI introduces new underlying language models and as the platform continues to scour the open internet and train itself with the data it finds.

Expect those B grades to inevitably improve. And as they do, expect the debates – over ethical use, impacts on employment, and educational best practice – to intensify.

THE INCUMBENTS ARE SCRAMBLING

As you can imagine, the public release of ChatGPT and the subsequent explosion in global interest is prompting some radical responses from industry stalwarts. Google likely has the most to lose, as chatbots represent an existential threat to the very future of search. The company has declared a “code red” and is actively reorganizing itself to develop AI-based technologies and hopefully not be outflanked in the process. It hurriedly released its Bard platform, but a public demo went awry when the Googlified bot mistakenly said that the James Webb Space Telescope had taken the first-ever photos of a planet outside of our solar system. Apprehensive investors panic-sold, and parent company Alphabet saw $100 billion in market value evaporate.

Microsoft is having somewhat of a better time after greenlighting an additional investment of $10 billion USD into OpenAI, and has since begun integrating the technology into its core products. Millions have signed up to access the remodeled Bing search engine – updated with an OpenAI-based chatbot originally code-named Sydney – and Microsoft has similar integration plans for Office, Teams, and Edge. 

Google has every reason to be scared.

MOVING BEYOND KNEE-JERK REACTIONS

To no one’s surprise, educators are worried about the potential for cheating – and some, including public school administrations in New York City and Seattle, have banned ChatGPT outright from school networks and devices. It’s a bit of a different story in higher education, where colleges and universities worry about compromising academic integrity, and instead of banning the tools outright are working on ways to change the way they teach and evaluate students’ work.

The academic experience holds significant lessons for businesses; avoiding the technology outright, is no longer a viable choice. Indeed, it never was. And while some roles within any given business may be disrupted out of existence, it’s significantly more likely that existing roles will be freed from administrative drudgework. New roles will be developed that allow workers to expand their productivity, increase their efficiency and expand their overall business value. 

Chatbots make great thought-starters and brainstorming aids. When faced with the proverbial blank screen, a creative knowledge worker can use a chatbot to drive the ideation process, get inspired, and refine ideas and concepts. It can help expose and explore new facets and angles and highlight new areas of research. We’ve been using search engines for decades to spark and drive the creative process, so chatbots are really little more than another, more expansive tool to add to our professional toolkit.

NOT SO FAST

Amid the hype around ChatGPT specifically and AI in general, we seem to be missing out on the technology’s fundamental weaknesses, that in their current state, they’re susceptible to churning out factually wrong, and sometimes dangerous, content. We forget that they are only as good as the data used to train them. And on that latter basis, it’s clear the engines are ingesting a lot of questionable data, and they aren’t sophisticated enough (yet) to tell legitimate content from outright misinformation.

The danger lies in what we ultimately do with what a chatbot spits out. If we blindly use it as is, we’re making ourselves unnecessarily vulnerable. If we take the output, then apply our own research and refinement to it, we’d likely catch and resolve any erroneous information. Knowledge workers have always brought their own sense of due diligence to their roles, and this does not change in the age of AI.

If anything, the advent of generative AI, like any revolutionary technology before it, allows writers, software developers, and other creatives to ditch the admin-heavy aspects of their roles and find new ways to create and distribute value. 

THE BOTTOM LINE

ChatGPT and other AI-based tools like it will rock our world, just as other groundbreaking new technologies have done throughout history. We owe it to ourselves to dive right in and figure out what it can and cannot do, so that we can extract maximum value from it – while at the same time, protecting ourselves from the inevitable risks. Sticking our heads in the sand is not the answer. 

The developers at STEP Software are already evaluating ChatGPT and similar generative AI tools to better understand how we can help further refine the software development process. If you’re still trying to figure out how generative AI can help your business, click here. We promise you we’ll continue to follow this generationally significant and disruptive technology, and we similarly promise you’ll always end up speaking with a real human being, if you want to know more.

Note: This article is part of our ongoing Creative Disruption series. In these articles, we explore what happens when technologies once seen as science-fiction become everyday reality. Check out the initial article – Creative Disruption – why we should embrace change, not fear it – for more, and click here if you’d like to suggest a future topic.

The post ChatGPT’s arrival reinforces the need to embrace – not ban – AI chatbots first appeared on STEP Software.

The Windows Local Security Authority Subsystem (LSASS) uses a type of virtual machine similar to the InPrivate Desktop sandbox. LSASS handles many of the security features of Windows. The reason it’s kept in a virtual environment is to protect it from malware. The “lsass.exe” process can be seen from the Windows Task Manager and must run to keep Windows from shutting down to the login screen. 

Passwords, password changes, hashes, and access tokens are all kept in Windows virtual machine to protect it from malware. Should a computer be compromised by malware, it’s unable to access critical data running on the operating system. Some attackers attempt to trick users into running a fraudulent executable that looks like the official exe. For example, “Lsass.exe” (capital “L” instead of lowercase “l”) is often distributed tricking users into running it as an official Windows application.

VIRTUAL MACHINES AND SANDBOX ENVIRONMENTS

Virtual machines (VMs) have long been a way for developers to host content that might be buggy or crash an operating system. Sandboxed environments cut off the rest of a network from the environment running the untrusted application. For instance, companies that analyze and develop applications to defend against malware or remove it from a system will use a sandbox environment for testing. You wouldn’t install malware on a machine that has access to the network, because the malware can spread across the network. Instead, these developers place malware and antivirus applications that remove it on a computer that has no access to the main network.

When an administrator sets up a virtual machine, it virtualizes the operating system, memory and computer resources. The underlying operating system is not affected, and administrators can be assured that anything running in the VM won’t be able to gain access to the underlying file system. The administrator can create and destroy the virtual machine as-needed, but destroying the VM also destroys any files stored within the sandbox. This is a benefit for some activities, such as testing and viewing results from running untrusted applications.

WHAT THE WINDOWS 10 SANDBOX VIRTUAL MACHINE OFFERS

The new Windows 10 VM sandbox will run similarly to a traditional VM environment with one exception: the new sandbox will have threads assigned a priority similar to the way a Windows application is given a priority. This priority system makes a Windows 10 machine more responsive should a higher priority application need resources currently in use by a low priority thread.

When a VM is running, it is assigned a priority that can be changed by the administrator. If the VM is given low priority, then an application with higher priority can “steal” the resources giving the high priority application better ability to stay functional even with a VM that uses too many resources.

InPrivate Desktop has some elements of a true virtual machine, but it also has features similar to Windows Containers. It still has a hardware boundary that separates VM resources from the host computer and virtualizes resources from the underlying machine. However, applications running on the same operating system as the local machine will share memory and disk resources, which is similar to Windows Containers.

Windows 10 Home users won’t see this new feature; it’s only available to Pro and Enterprise Windows 10 editions. 

WHAT THIS MEANS FOR DEVELOPERS

Developers work with testing environments usually in a VM or possibly in the newer container architecture. With the new Windows 10 feature, developers can more easily test applications directly on their desktop instead of moving files to a server to test. They can still run applications in debug mode to view any errors, but compiled versions of software can be run in the VM. This saves time for developers that make several version changes and upgrades to corporate software.

This new feature also gives end users a better way to install untrusted applications. Should the user prefer to install an application but is unsure of its validity, the user can run it in the VM to identify how it interacts with a computer. Should the application be identified as malware, the VM can be destroyed and the application sandboxed from the main operating system. Microsoft’s feature gives Windows 10 much better safety from malware should this get released to all editions of the operating system.

The post Microsoft Announces a Sandbox Environment for Windows 10 first appeared on STEP Software.