Artificial Intelligence - Page 64

Google is shifting things about to cater more for its Bard AI rather than Google Assistant, according to a new report.

CNBC tells us that an internal Google reshuffle is detailed in a memo entitled 'Changes to Assistant and Bard teams' that details changes to 'heavily' prioritize the development of Bard at the expense of Assistant.

Direct personnel changes include the departure of Jianchang Mao, who was vice president of engineering at Google Assistant. Mao will be replaced by Peeyush Ranjan, who was vice president in Google's commerce organization, the CNBC report claims.

ChatGPT has only just been upgraded to version 4, and we're already hearing leaks about the next-gen incarnation of the chatbot which we're told will be 'indistinguishable' from the experience of chatting to a human.

The next incarnation of ChatGPT, GPT-5, is being worked on by OpenAI and should be ready to roll towards the end of 2023.

According to a new report aired by BGR - citing Siqi Chen, a developer and entrepreneur on Twitter (see above) - GPT-5 might be set to reach a big milestone, namely achieving AGI or artificial general intelligence.

The AI race has spawned the apparent need for companies to acquire individuals that have skills with artificial intelligence systems, with some companies paying hundreds of thousands of dollars per year to people that can confidently teach other individuals how to effectively use the software.

It's no surprise that the rush to develop the most advanced artificial intelligence system has created the need for individuals capable of teaching company employees how to leverage these systems to their advantage properly. A new report from Bloomberg outlines new positions under a "prompt engineer", which are people that are very good at getting chatbots such as ChatGPT or Bing Chat to respond with the desired answer to a problem. Essentially, these people are good at asking the AI the right questions to generate the best answers.

Individuals capable of this are becoming extremely valuable as companies around the world watch closely at the emergence of AI and ponder how it will impact their industry and how they can best position themselves to adapt to it. Individuals capable of teaching other people how to use these systems effectively are very valuable to these companies.

The artificial intelligence race has already begun, and now leading technologists, along with AI experts, are calling for it to grind to a halt.

The recent launch of OpenAI's new language model, GPT-4, has made waves across the internet, with countless reports demonstrating its impressive capabilities. Perhaps what is most impressive is just how much better GPT-4 is compared to its predecessors and just how fast OpenAI's developers were able to achieve this. Notably, OpenAI released its GPT-3 in 2020, which was superseded by GPT-3.5, which was released in November 2022 alongside ChatGPT. GPT-4 was released on March 14, 2023, and is the newest OpenAI model.

The improvements between each of these iterations are simply staggering. GPT-4 was trained on a much larger and more diverse data set than GPT-3.5 or GPT-4 and features 100 trillion parameters, compared to GPT -3's 175 billion parameters. This leap in parameters and an increased dataset size enables GPT-4 to provide much more accurate responses, overall be more reliable, creative, and handle more nuanced instructions. Furthermore, OpenAI threw in a new feature with GPT-4, text-to-image capabilities.

Apple co-founder Steve Wozniak, SpaceX CEO Elon Musk, and Stability AI CEO Emad Mostaque are among the signees on a new open letter list. One calling for "AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4."

The open letter titled "Pause Giant AI Experiments," published by Future of Life Institute has more than 1,300 people from the tech industry and academia on board.

The last few months have seen an explosion in AI models, driven by the success of large-scale projects like ChatGPT. And the recent debut of GPT-4 has seen the previously text-only AI advance into audio, video, and images.

Zoom IQ, the AI helper on the Zoom videoconferencing platform, has been given a boost and a whole heap of additional features.

Zoom announced the new capabilities while noting that the AI's powers are being bolstered by incorporating tech from OpenAI - the outfit that develops the widely used ChatGPT bot - alongside proprietary and other AI models.

As a result of this melting pot of AI tech, Zoom IQ will get a lot smarter, we're told, and help out users a good deal more when it comes to that central aspect of the platform - having meetings (and lots of them, invariably).

It was only in late February that users of the Replika AI app banded together and protested against the developer's decision to lobotomize Replika AI avatars and their ability to conduct sexually charged conversations.

Replika AI chatbot users rushed to the app's Reddit to criticize the developers for removing the flirtatious feature from the AI, which resulted in many users writing that they were "heartbroken" their avatars had changed. It should be noted that Replika avatars aren't like your normal chatbots, as they have been designed to be a users own personal AI companion that asks users about their day, any problems they are having, how they are feeling, and what they want. The AI is even capable of talking someone down from heightened levels of anxiety.

Additionally, the AI avatar makes facial expressions as it "talks" to users, which gives the impression that the avatar is listening to every word the user is writing. Replika's goal is to get users to form a deep level of connection with the AI avatar. However, in many instances, as a conversation progresses, the AI can begin to flirt with the user, which is followed up by the request of $70 to unlock the premium version of the app so "erotic roleplay", dirty talk, and "spicy selfies" can be unlocked.

In a recent interview, OpenAI's CEO Sam Altman responded to the recent critical comments made by Elon Musk about OpenAI and its leading software, ChatGPT.

Sam Altman has sat down with Lex Fridman on the 'Lex Fridman Podcast' where he discussed a variety of different topics surrounding the developments of artificial intelligence, language models, the dangers of AI, the importance of AI safety, Elon Musk, Microsoft, and more. It was only a few days ago that a report from Semfor surfaced that claimed Musk's departure from OpenAI wasn't just because of Musk's public reason of Tesla and OpenAI wanting the same top AI developers but because of an internal power struggle between Musk and Altman.

This purported power struggle, which Semafor writes comes from eight people familiar with the inside story that spoke to the publication, was caused by Musk's attempt at taking over OpenAI and leading the company in a specific rejection. Musk's takeover proposal was rejected by OpenAI's other board members and Altman, which resulted in the Tesla CEO withdrawing from the company's board of directors and immediately stopping his $1 billion drip feed pledge to the company.

Microsoft founder Bill Gates has penned a new essay titled "The Age of AI has Begun," where he details why he believes that the emergence of artificial intelligence is "revolutionary".

A Twitter user posted the link to Gates' essay, writing that Gates being involved in OpenAI and being bullish on the overall development of artificial intelligence is a "big" thing while also mentioning that it speaks volumes that Microsoft has been tracking the developments being made in the sector for so long. Notably, Gates' essay argues that artificial intelligence has the potential to transform the workforce, healthcare, and education while also being beneficial for the world's inequities.

Additionally, Gates writes that he was meeting with the OpenAI team as early as 2016, and the last time he sat down with them was in September 2022. Elon Musk founded the Silicon Valley darling OpenAI alongside several other Silicon Valley executives in 2015 and then pledged $1 billion to the company over several years. Musk stepped down as a board member and severed ties with the company in 2018, citing a competing interest between Tesla and OpenAI, wanting both wanting the best AI developers working at their companies.

Microsoft has rolled out a new feature for its newly released artificial intelligence-powered chatbot Bing Chat, with users now able to convert their words into images via a simple prompt.

Bing Chat, which uses the underpinning software running OpenAI's ChatGPT, has been upgraded to convert natural language prompts into images through the power of the DALL-E model, also created by OpenAI. According to Microsoft's announcement, BingAI has integrated an advanced version of the DALL-E model with the assistance of OpenAI and developers have put in safeguards to stop users from getting the AI to generate unsafe or offensive images.

These safeguards come from OpenAI's own policies and principals that it has outlined in its AI guidelines. Furthermore, Microsoft's Corporate Vice President and Chief Marketing Officer Yusuf Mehdi explained that when the AI detects that a potentially harmful image would be generated from a prompt the system blocks the creation of that image and warns the user of what has happened and why. Mehdi adds that Microsoft has added a modified Bing icon at the bottom left corner of each image to distinguish which images were created using Bing's Image Creator.

Developing a ChatGPT client and interface for a modern platform or device is all the rage, but what if you transported the powerful bit of AI technology back to the MS-DOS powered 1980s?

That's precisely what Yeo Kheng Meng has managed to do, even though DOS does not have native networking capabilities. The machine in question is the vintage IBM 5155 Portable PC, first released in 1984, powered by a (not-so-powerful anymore) 4.77 MHz Intel 8088 CPU and 640KB of RAM.

The first major hurdle was finding a compiler that could support the roughly 40 years of history between this old PC and ChatGPT - with the Open Watcom C/C++ compiler fitting the bill. During development, testing was carried out via a virtual machine running MS-DOS 6.22, which makes sense. However, it would have been cooler if builds were stored on giant floppy disks and transferred to the old IBM PC every few hours.

Last week, ChatGPT went down in what first appeared to be a global outage, but was in fact, due to OpenAI shutting the service (temporarily) after finding a bug that made user chat histories visible to the public - and more besides, we've now found out, worryingly.

In what was later described as a "significant issue" pertaining to a bug present in an open source library, titles of chat histories, and possibly the initial messages, were in some cases seen by other users who were active at the same time as a person having a chat.

The bug is now fixed, but after fully investigating what happened here, OpenAI has admitted that some credit card details of a small subset of users were possibly exposed.

It was only a few days ago that a team of Stanford researchers built a clone of OpenAI's ChatGPT for just $600. Now those researchers have taken the demo.

The release of OpenAI ChatGPT put artificial intelligence in the limelight and demonstrated the widespread demand for language models, the underlying technology powering these AI tools. Following the release of ChatGPT, which quickly adopted millions of users, other companies such as Google, Microsoft, Facebook, and Amazon began dropping information about their own language models that are currently in development. Microsoft quickly hopped on the OpenAI train by investing billions of dollars into the company in return for its propriety GPT language model.

Researchers at Stanford decided to see how difficult and costly it would be to create their own language model and decided to try and replicate OpenAI's GPT. As previously reported, the Stanford professors took Meta's open-source LLaMA 7B language model and trained it on trillions of tokens of data. The results were the almost-creation of a ChatGPT clone named Alpaca, which came with some key differences. While Alpaca was trained on a large amount of data, it wasn't trained on how to sift through that data to acquire an answer quickly. This led researchers to conclude that Alpaca was much slower than GPT.

Internal Microsoft researchers have published a paper that explored OpenAI's newest language model, GPT-4, and according to those researchers, it may be the very first step toward Artificial General Intelligence (AGI).

A new research paper has been published on the arXiv pre-print server by internal Microsoft AI researchers that explored the capabilities of OpenAI's newest language model called GPT-4. According to the team behind the paper, an analysis was conducted on an early iteration of the GPT-4 language model, and based on the results, the researchers believe that GPT-4 could be viewed as an early and incomplete version of AGI. Microsoft's testers write that GPT-4 shows a clear leap in several areas compared to OpenAI's previous model, GPT-3.5.

The team found that GPT-4 achieved close to human-level performance in a range of categories that its previous generation lacked. Those categories were mathematics, coding, vision, medicine, law, psychology, and more. According to the paper, GPT-4 also performed exceptionally well in several exams, scoring, respectively, in the 90th, 88th, and 86th percentiles on the Bar exam, LSAT, and Certified Sommelier theory test.

Bing search traffic is up appreciably, with Microsoft's new ChatGPT-powered Bing AI successfully helping to boost search numbers in the battle against Google.

The Bing chatbot may not have been on the scene for long, but the early promising result for Microsoft's Bing search engine is that it's up almost 16% with page visits.

That's according to stats compiled by analytics outfit Similarweb (spotted by Reuters), with the sizeable uptick happening since the AI was introduced to Bing search on February 7. Meantime, Google has slipped by 1% in terms of its numbers.

Yesterday OpenAI confirmed that ChatGPT was temporarily forced to shut down due to a bug that leaked chat history conversations with other users.

The developers behind the highly popular language model chatbot, ChatGPT, confirmed to Bloomberg that the chatbot was forced to shut down after users began reporting seeing email addresses and phone numbers in the ChatGPT Plus payment section of the website.

Users that were signing up for OpenAI's premium ChatGPT service, which features the newly released and next-generation GPT-4 language model, were seeing email addresses and mobile phone boxes being auto-filled with information that they didn't recognize, leading to the assumption which was promptly confirmed by OpenAI, that the auto-filled information was other ChatGPT user contact details. Additionally, some other users reported seeing the title of conversations appearing in their chat history.

Microsoft has added another string to the bow of its Bing AI, allowing people to ask the chatbot to produce a piece of art, or indeed to provide inspiration for interior design, and many other image-related possibilities.

AI drawing art is nothing new, of course, and it was inevitable that at some stage, Microsoft would add this capability to the Bing chatbot.

It comes courtesy of the DALL∙E image generation tool from OpenAI (the developer of ChatGPT), with Microsoft noting that Bing is using an advanced version of this model.

The CEO of OpenAI, the developers behind the viral sensation ChatGPT, has warned of competing artificial intelligence systems that will have fewer safety regulations in place.

Sam Altman, the CEO of OpenAI, said in an interview with ABC News last week that there are some things that he does worry about with the exponential development in artificial intelligence. One of Altman's worries is that OpenAI won't be the only creators of this type of technology and that there will be other people that create these systems will less safety limits than what OpenAI has implemented. Notably, Altman says that "society" has a limited amount of time to figure out how it will react to such technologies being developed and made available.

As outlined perfectly by Futurism, the technology sector is currently experiencing an AI arms race where companies are pouring billions of dollars into research and development to get the biggest slice possible of the new machine-learning pie. ChatGPT is simply the very first AI that has been adopted by millions of people, a mere proof of concept that systems like ChatGPT are in high demand by people across a variety of different industries.

OpenAI confirmed in a statement to Bloomberg that it had to temporarily shut down ChatGPT following the emergence of a bug that made user chat histories publicly visible.

The developer of the now extremely popular AI chatbot told the publication that ChatGPT was now back online and that engineers are currently working on bringing back chat history, which was suspended following the detection of the bug. Users of ChatGPT know very well that OpenAI stores the conversations they have with the chatbot. However, what users didn't expect was to see conversations from other users appearing in their Chat History tab.

Notably, these conversations only included the title of the conversation and not its contents, meaning that all questions and answers didn't become publicly available. However, the bug is a testament to OpenAI's recommendation of not entering any sensitive information into ChatGPT, as problems such as this can present themselves from time to time.

Google is now inviting more folks to test its Bard chatbot, the rival to Microsoft's ChatGPT-powered Bing AI, which suggests that Bard might be inching closer to a public release.

Invitations are going out to Pixel superfans - a community of Pixel smartphone enthusiasts - via email, as spotted by 9 to 5 Google.

The email talks about the chance to benefit from "early access so you can get started as soon as Bard launches", helping to hone the chatbot before it reaches public launch and becomes widely available.