Monthly Archives: May 2023

Neuralink, the brain chip firm founded by Elon Musk, has secured a momentous achievement as the US Food and Drug Administration (FDA) grants approval for its first human clinical trial. This milestone follows earlier struggles the company faced in obtaining regulatory clearance, positioning Neuralink at the forefront of the race to develop a revolutionary brain-computer interface.

In a tweet, Neuralink expressed that the FDA’s approval represents a crucial initial step toward a future where their technology can assist countless individuals. Although specific details of the study have not been disclosed, the company stated that recruitment has not commenced and promised to share more information soon.

Neuralink’s primary objective is to establish a groundbreaking method for direct communication between computers and the human brain. By accomplishing this, the company aims to address complex neurological conditions that have thus far eluded effective medical interventions. These conditions encompass a wide range, from paralysis and blindness to depression and schizophrenia.

The proposed procedure involves implanting a minute, hermetically sealed chip directly into the patient’s brain. This chip is connected to 1,024 tiny electrodes, thinner than a human hair, and powered by a rechargeable wireless battery. Through this interface, the chip can both transmit and receive signals from an external computer.

Safety concerns surrounding Neuralink’s technology encompass three aspects: short-term physical risks, long-term medical implications, and ethical considerations. Brain surgeries inherently carry the risk of physical harm and potential rejection by the host organism. While the company has conducted extensive (albeit controversial) testing on animals, a complaint filed earlier this year highlighted alleged mistreatment of macaque monkeys involved in the technology’s evaluation.

However, the FDA’s approval for human testing suggests that Neuralink has overcome some of these challenges. Yet, more profound concerns arise regarding the long-term effects of having such a device operating within the immensely intricate human brain—an organ that humanity is only beginning to comprehend. Currently, there is no data on potential harm, but human testing will provide crucial insights for the development of similar products.

Ethical considerations related to Neuralink’s technology are subjective and contentious. The field of brain-computer interfaces raises concerns about data protection, potential applications, and the controversial concept of human enhancement. This includes augmenting cognitive abilities, sensory perception, or physical capabilities beyond current human limits. Consequently, calls for stringent regulation of this sector are anticipated.

Neuralink was established in California by Elon Musk and a team of seven scientists and engineers in 2016. Over time, most of the co-founders have departed, leaving only two at the company. Musk’s involvement brings an entrepreneurial approach and substantial publicity to an issue that many primarily consider scientific in nature. Being one of the most recognizable and controversial figures globally, Musk’s association with Neuralink ensures widespread attention and discourse surrounding the venture.

Several other companies are actively pursuing similar endeavors, with some reportedly making more progress than Neuralink. For instance, Black Rock Neuro Tech is also working on a brain implant technology. Additionally, Meta, the parent company of Facebook, is rumored to be developing wearable tech that enables typing through neural signals. Recently, Swiss scientists achieved a breakthrough in assisting a paralyzed individual to walk again. The field is rapidly expanding with numerous players vying for success.

While Neuralink’s association with Elon Musk guarantees substantial publicity, the ultimate victors in this race will be determined by which technology proves to be the most effective and can demonstrate safety to regulators and the market.

In an anticipated move, Google has introduced the beta version of Magic Compose, a new feature for its Messages app that leverages AI to aid in writing text messages. However, this exciting addition comes with a significant caveat. Android Police has brought to light that Magic Compose will send up to “20 previous messages” to Google’s servers, even when using RCS with end-to-end encryption (E2EE).

data:image/gif;base64,R0lGODlhAQABAAAAACH5BAEKAAEALAAAAAABAAEAAAICTAEAOw==

The conditions surrounding Magic Compose are outlined on Google’s support page, clarifying that the feature transmits these messages, including emojis, reactions, and URLs, to its servers to enable AI-driven suggestions. The company assures users that it will not forward any messages containing attachments, voice messages, or images. However, it mentions that image captions and voice transcriptions might be sent.

Google implemented end-to-end encryption on its app in 2020 and extended it to group chats last year. Activating E2EE ensures that not even Google can access users’ messages. Despite using Magic Compose alongside E2EE, Google maintains that it cannot read the messages.

Google spokesperson Justin Rende clarified to The Verge that the “conversation data used by Magic Compose is not retained” and that “suggested response outputs are not retained once they’ve been provided to the user.” Furthermore, when users disable Magic Compose, Google will no longer transmit their messages to its servers.

Magic Compose is just one of the many AI-powered features showcased by Google during its I/O event earlier this month. This feature enables users to respond to text messages using stylized, context-based suggested responses. It is currently being rolled out to users participating in the Google Messages beta program.

Once the feature becomes available, users will notice a chat bubble next to the app’s message composer. They can choose a suggested response and further customize the text using preset styles such as “chill,” “excited,” or “Shakespeare.” Presently, Magic Compose appears to be limited to RCS messages, and there is no information on when it will support SMS/MMS.

In a similar move, Microsoft introduced a comparable feature in its keyboard app, SwiftKey. This functionality enables users to compose text messages and emails, modify the tone, format, and length of suggested messages, and even includes a Bing icon within the app’s toolbar.

While Magic Compose presents an exciting development for Google Messages users, concerns about privacy and the transmission of messages to Google’s servers have been raised. Users should weigh the convenience of AI assistance against their privacy preferences before deciding to utilize this feature.

TikTok, the popular social media platform, is currently conducting tests on an innovative AI chatbot named Tako, which has the potential to transform the way users search and navigate within the app. Screenshots of the feature, shared with The Verge, reveal that Tako can recommend videos based on user inquiries.

According to Daniel Buchuk from Watchful Technologies, a firm specializing in identifying upcoming app changes for Fortune 500 companies, if TikTok decides to release Tako on a wider scale, it could “radically change search and navigation” within the platform. The precise AI model employed by TikTok to power Tako remains undisclosed.

In the shared screenshots, Tako is positioned above the TikTok profile icon, adjacent to a video. Tapping on Tako opens a chat interface, where the bot appears capable of providing answers to a broad range of queries. The extent of Tako’s capabilities and its proficiency in understanding user intent are yet to be fully revealed.

https://platform.twitter.com/embed/Tweet.html?dnt=true&embedId=twitter-widget-0&features=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%3D%3D&frame=false&hideCard=false&hideThread=false&id=1661718779024531456&lang=en&origin=https%3A%2F%2Fsoftwareterritory.com%2Ftiktok-testing-ai-chatbot-tako-to-revolutionize-search-and-navigation-within-the-app%2F&sessionId=b9ea5df35b42978ab143399631d3cab9228b06d8&theme=light&widgetsVersion=aaf4084522e3a%3A1674595607486&width=550px

Tako is designed to offer suggested prompts that aid users in initiating conversations with the chatbot. Buchuk explains that when watching food videos and requesting a recipe, Tako will present related TikTok videos alongside a list of suggestions in response. Another screenshot displays a Tako-suggested prompt asking about the significance of King Charles III’s coronation.

Zachary Kizer, a spokesperson for TikTok, emphasized that the chatbot is currently in the early stages as a “limited experiment” and is not accessible to users in North America or Europe. The test is exclusively being conducted in the Philippines, according to a tweet from the company.

Kizer further added, “We’re always exploring new technologies that add value to our community. In select markets, we’re testing new ways to power search and discovery on TikTok, and we look forward to learning from our community as we continue to create a safe place that entertains, inspires creativity, and drives culture.”

Notably, TikTok recently filed a trademark application for “chatbot software” under the name Tako, suggesting that the company is preparing for a broader release of the AI chatbot. TikTok follows in the footsteps of other social media platforms, such as Snapchat, which made its My AI bot freely available to all users in April. Mark Zuckerberg has also expressed his intent to introduce AI agents to billions of people, showcasing the growing trend of chatbot integration in social media.

The potential introduction of Tako as a widely accessible AI chatbot on TikTok signifies the platform’s dedication to enhancing user experiences through advanced technologies. As the testing phase continues, users eagerly await the arrival of this innovative feature, which holds the promise of transforming the way they interact with TikTok’s vast content library.

Adobe Photoshop is revolutionizing image editing with its latest feature called Generative Fill, harnessing the power of AI to extend images and manipulate objects based on text prompts. This cutting-edge tool, utilizing Adobe’s AI image generator Firefly, was previously available as a web-only beta and is now launching in beta within Photoshop, with a full release expected later this year.

Generative Fill seamlessly integrates into the Photoshop workflow, operating within individual layers of an image file. Whether users want to expand image borders (outpainting) or create new objects, the tool presents three options for selection. When used for outpainting, the system can generate image extensions automatically, but it performs even better with user-guided instructions. This functionality bears resemblance to Photoshop’s existing Content-Aware Fill feature, offering users enhanced control over the editing process.

Although I haven’t personally tested Generative Fill, I witnessed a live demonstration that showcased its impressive capabilities alongside its limitations. While some generated objects, such as cars and puddles, appeared unnatural within the image, the tool excelled at handling backgrounds and filling empty spaces. In certain instances, it even managed to incorporate visual elements from the original photograph, mimicking light sources and creating reflections in generated water.

For creators familiar with AI image generation tools, such feats may not be entirely surprising. However, the integration of this technology into mainstream applications like Photoshop brings these capabilities to a broader audience.

In addition to its functionality, Firefly’s training data stands as another crucial aspect. Adobe assures that the model is exclusively trained on content for which the company possesses the necessary rights, including Adobe Stock images, openly licensed content, and copyright-restricted content. In theory, this ensures that anything created using the Generative Fill feature is safe for commercial use, offering a sense of security to creatives and agencies who may have been hesitant to utilize AI tools due to potential legal implications.

Generative Fill also incorporates Content Credentials, a system likened to a “nutrition label” that attaches attribution data to images when shared online. This information informs viewers whether the content was created or modified using AI. To verify the Content Credentials of an image, users can inspect it via verify.contentauthenticity.org, which provides a comprehensive overview of relevant information.

Ashley Still, Senior Vice President of Digital Media at Adobe, emphasized the impact of integrating Firefly directly into creative workflows, stating, “By integrating Firefly directly into workflows as a creative co-pilot, Adobe is accelerating ideation, exploration, and production for all of our customers. Generative Fill combines the speed and ease of generative AI with the power and precision of Photoshop, empowering customers to bring their visions to life at the speed of their imaginations.”

While Generative Fill is not yet available in the full release of Photoshop, users can experience it firsthand by downloading the desktop beta app or accessing it as a module within the Firefly beta app. Adobe plans to introduce the feature to the public Photoshop app in the second half of 2023.

Adobe has been progressively incorporating AI-powered tools into its products, with Sensei, another of Adobe’s AI models, driving features like higher-quality object selections in Photoshop. Firefly is already utilized in Adobe Illustrator for vector-based image recoloring, and Adobe intends to integrate Firefly with Adobe Express, a cloud-based design platform that competes with services like Canva, although no official release date has been confirmed.

Meta, the parent company of Facebook, has been issued a €1.2bn (£1bn) fine by Ireland’s Data Protection Commission (DPC) for mishandling the transfer of people’s data between Europe and the United States. This represents the largest fine ever imposed under the EU’s General Data Protection Regulation (GDPR) privacy law. The fine is a result of Meta’s alleged violation of GDPR rules that govern the transfer of user data outside of the EU. Meta has expressed its intention to appeal the ruling, citing it as “unjustified and unnecessary.”

Concerns have emerged regarding the use of standard contractual clauses (SCCs) in moving European Union data to the US. These legally binding contracts, formulated by the European Commission, contain safeguards to ensure the continued protection of personal data during its transfer outside of Europe. However, there are apprehensions that these data flows may still expose Europeans to weaker privacy laws in the US, and that US intelligence agencies could potentially access the data.

It is important to note that this decision does not directly affect Facebook in the UK. The Information Commissioner’s Office clarified that the ruling “does not apply in the UK,” although it acknowledged the decision and stated that it would review the details in due course.

Meta argues that the wide application of SCCs makes the imposed fine unjust. Facebook’s president, Nick Clegg, expressed disappointment, stating, “We are therefore disappointed to have been singled out when using the same legal mechanism as thousands of other companies looking to provide services in Europe. This decision is flawed, unjustified, and sets a dangerous precedent for the countless other companies transferring data between the EU and US.”

While Meta contests the ruling, privacy groups have welcomed it as a significant precedent. Caitlin Fennessy, of the International Association of Privacy Professionals, commented, “The size of this record-breaking fine is matched by the significance of the signal it sends. Today’s decision signals that companies have a whole lot of risk on the table.” The ruling could lead EU companies to demand that US partners store data within Europe or consider switching to domestic alternatives, Fennessy added.

The decade-long battle surrounding the legality of transferring EU data to the US traces back to Edward Snowden’s 2013 disclosure of US intelligence agencies repeatedly accessing individuals’ information through technology companies like Facebook and Google. Austrian privacy campaigner Max Schrems initiated a legal challenge against Facebook, claiming a failure to protect his privacy rights, which sparked the ongoing dispute. The European Court of Justice (ECJ), Europe’s highest court, has consistently maintained that US surveillance laws lack sufficient safeguards to protect Europeans’ information.

In 2020, the ECJ invalidated an EU-to-US data transfer agreement, although it allowed companies to use SCCs if they ensured an “adequate level of data protection” during transfers to other third countries. It is this criterion that Meta has been found to have failed.

In response to the €1.2bn fine, Max Schrems expressed satisfaction after a decade of litigation but suggested that the penalty could have been even higher. He emphasized that Meta would need to fundamentally restructure its systems unless US surveillance laws were rectified.

Despite the record-breaking fine, experts believe that Meta’s privacy practices are unlikely to undergo significant changes. Johnny Ryan, a senior fellow at the Irish Council for Civil Liberties, remarked, “A billion-euro parking ticket is of no consequence to a company that earns many more billions by parking illegally.”

The US has recently updated its internal legal protections to provide the EU with greater assurances that American intelligence agencies will adhere to new rules governing data access. In 2021, Amazon faced a similar fine for breaching the EU’s privacy standards. The DPC in Ireland has also fined WhatsApp, another business owned by Meta, for violating data protection regulations.

According to a report from Bloomberg News on Friday, Meta Platforms’ Instagram is gearing up to introduce a new text-based app that will directly rival Twitter. The app, which has been in development and testing, may potentially debut as early as June, sources familiar with the matter have revealed.

Influencers and selected creators have been chosen to participate in the testing phase of the product, as stated in the report. However, when contacted for comment, Meta Platforms did not immediately respond to a request from Reuters.

Insights from industry insiders suggest that Meta Platforms has been reaching out to talent agencies and celebrities, gauging their interest in trying out an early version of the app. This upcoming app will be integrated with Instagram but will also feature compatibility with other platforms like Mastodon, according to a newsletter by Alex Heath.

Lia Haberman, a social and influencer marketing instructor at UCLA, also shared details about the app. According to Haberman, the new app has been discreetly made available to select creators for several months now. While it is separate from Instagram, it will allow users to connect their accounts seamlessly.

A screenshot shared by Haberman provides a glimpse into the app’s features. Users will be able to connect with friends through texts, shared links, photos, and videos. Additionally, fans will have the opportunity to easily join the accounts of their favorite influencers and creators from Instagram with just a single tap.

In terms of security, the app is being updated to include robust measures. Accounts blocked on Instagram will soon carry over to this new app, ensuring privacy for its users. Furthermore, efforts are underway to ensure compatibility with Mastodon, Twitter’s competitor platform.

The release of this text-based app by Instagram, under the umbrella of Meta Platforms, is expected to intensify the competition between social media giants, particularly in the realm of text-based communication and influencer engagement. As June approaches, anticipation grows among users and industry observers, eager to see how this new app will impact the social media landscape.

BT Group, a leading UK telecom company, has announced its ambitious plan to reduce its workforce by up to 55,000 jobs in the next five to seven years. The company aims to leverage technology, including artificial intelligence (AI), to streamline operations, cut costs, and simplify its business structure.

A Smaller Workforce and Cost Reductions

BT’s total workforce, which currently stands at 130,000 employees and contractors, will decrease to between 75,000 and 90,000 by 2028-2030, according to the company’s recent announcement. CEO Philip Jansen emphasized the goal of relying on a smaller workforce and achieving a significantly reduced cost base through aggressive digitization efforts and organizational restructuring.

Embracing Digitization and Automation

During a call with analysts, Jansen highlighted that approximately 10,000 positions would be replaced by digitization and automation. The CEO expressed confidence in AI’s transformative power, stating that it would enable BT to deliver seamless customer service. AI-powered chatbots, such as “Amy,” have already proven effective in handling customer queries. BT also expressed interest in exploring new products and services that could emerge from generative AI and large language models.

Fewer Personnel Needed for Network Servicing and Repairs

Jansen noted that advancements in digital networks would result in a decreased demand for personnel involved in servicing and repairs. Modern networks experience fewer issues and can be repaired more efficiently than older infrastructure, leading to a reduced labor requirement in these areas.

Telecom Industry Challenges and BT’s Cost-Cutting Measures

The telecom industry has faced significant challenges, with traditional players struggling against competition from tech giants like Apple and Google. McKinsey, a leading consultancy, has pointed out that European telecom companies have underperformed compared to their US counterparts, delivering lower returns to shareholders over the past decade.

BT has already implemented measures to reduce costs, slashing £2.1 billion ($2.6 billion) since April 2020. Although the planned job cuts may seem drastic, industry analysts suggest they are not surprising given the increasing costs and slim margins within the broader telecom business. Matt Britzman, an equity analyst at broker Hargreaves Lansdown, emphasized that once BT completes its broadband and 5G network expansions, the focus will shift towards monetizing the existing infrastructure and leveraging new technologies.

Financial Performance and Investor Concerns

In its financial report for the year ending in March, BT disclosed a 1% decline in revenue to £20.7 billion ($25.8 billion). While Openreach, the company’s fiber broadband network, experienced growth, declines were observed in other business sectors. However, adjusted earnings saw a 5% increase, reaching £7.9 billion ($9.8 billion).

BT’s announcement of job cuts led to an 8% drop in its London-listed shares as investors expressed concerns about increased expenditure leading to a decrease in free cash flow.

Navigating a Transformative Phase in the Telecom Industry

The telecom industry is currently undergoing a transformative phase as companies like BT and Vodafone adapt to the evolving landscape by embracing technology and making strategic workforce adjustments. The coming years will test their ability to navigate these challenges and remain competitive in an increasingly digital world.

At the highly anticipated Google I/O 2023 event, the tech giant introduced its latest groundbreaking advancement in the field of artificial intelligence: PaLM 2. Serving as the foundation for several upcoming Google products, including Google Generative AI Search, Duet AI in Google Docs and Gmail, and Google Bard, PaLM 2 is a remarkable general-purpose large language model. This article aims to provide an in-depth overview of the PaLM 2 AI model, exploring its features, capabilities, and how it compares to its predecessor, GPT-4.

What is Google’s PaLM 2 AI Model?

PaLM 2 represents the most recent release of Google’s Large Language Model (LLM). This highly capable model excels in advanced reasoning, coding, and mathematics, making it a versatile tool. Supporting over 100 languages, PaLM 2 is a successor to the earlier Pathways Language Model (PaLM) that made its debut in 2022. Compared to its predecessor, PaLM 2 has undergone significant advancements. While the original PaLM was trained on an impressive 540 billion parameters, the newer PaLM 2 model is smaller in size yet faster and more efficient.

Understanding the PaLM 2 AI Model

Although Google has not explicitly disclosed the parameter size of PaLM 2 in its 92-page technical report, reports from TechCrunch suggest that one of the PaLM 2 models has been trained on 14.7 billion parameters. This parameter count is considerably lower than that of PaLM 1 and other competing models. Speculation on Twitter by researchers suggests that the largest PaLM 2 model could be trained on around 100 billion parameters, still significantly smaller than the parameter count of GPT-4.

For comparison, OpenAI’s GPT-4 model is reportedly trained on a staggering 1 trillion parameters, showcasing the magnitude of its scale. GPT-4 is approximately ten times larger than PaLM 2 in terms of parameters.

The Smaller Footprint of PaLM 2

Google emphasizes that size isn’t always an indication of superiority. In its official blog post, the company highlights the importance of research creativity in creating exceptional models. While the specific research creativity employed in PaLM 2 remains undisclosed, it is likely that Google has utilized techniques such as Reinforcement Learning from Human Feedback (RLHF), compute-optimal scaling, LoRA (Low-Rank Adaptation), instruction tuning, and high-quality datasets. These techniques contribute to achieving impressive results with a relatively smaller model.

Highlight Features of PaLM 2

PaLM 2 boasts a range of features that set it apart. Firstly, the model excels in common sense reasoning, with Google affirming that PaLM 2’s reasoning capabilities rival those of GPT-4. In various reasoning tests such as the WinoGrande commonsense test, PaLM 2 achieved a score of 90.2, outperforming GPT-4’s score of 87.5. In other tests like the ARC-C test, GPT-4 achieved a slightly higher score of 96.3 compared to PaLM 2’s 95.1. PaLM 2 also outperforms GPT-4 in reasoning evaluations including DROP, StrategyQA, and CSQA.

Additionally, PaLM 2’s multilingual capabilities enable it to understand idioms, poems, nuanced texts, and riddles in different languages. It surpasses literal word meanings and comprehends the ambiguous and figurative aspects of language. Pre-training on parallel multilingual texts and a corpus of high-quality multilingual data enhance PaLM