The battle between Open Source and closed systems heats up in both Linux and Artificial Intelligence. J. P Morgan gets fined big bucks for deleting data and a robot dog with a flamethrower – what could possibly go wrong with that.
 |
 |
 |
These and more top tech news stories from Hashtag Trending and Tech News Day. I’m your host Jim Love, CIO of IT World Canada and Tech News Day in the US.
In the world of artificial intelligence, a battle is brewing between open-source and closed AI models. Open-source models, which allow anyone to view and adapt the code, are gaining traction and enabling companies to have their own in house AI and for smaller players to compete with market leaders like ChatGPT.
It’s not just startups. Tech giants like Meta, and even governments are all showing interest in open-source AI. Meta’s recent release of its 65 billion-parameter LLaMA foundation model has made building high-quality AI models accessible to almost anyone. And more companies, like Mosaic, are now releasing their own open-source models that claim to outperform original models like GPT-3.
However, not everyone is on board with the open-source movement. OpenAI and Google, early pioneers in generative AI, are keeping their models closed to maintain their competitive edge. OpenAI’s ChatGPT, despite the company’s name, is also a closed model.
But the debate over open versus closed models is not just about competition. It’s also about balancing accountability, innovation and possibly safety.
On the positive side, Hugging Face founder Clement Delangue testified before Congress that open-source models prevent “black-box systems” and fosters innovation across the economy.
But on the negative side, there are risks that cannot be ignored. Senators in the US have expressed concerns about the potential misuse of open-source models. They’ve asked Meta for proof of efforts to mitigate these risks. But as we noted, there are dozens of Open Source AI models out there. Axios reported 37 of them in one article. So going after Meta might be trying to close the door after the horses have left the barn.
In the end, the choice between open and closed models may not be clear-cut. Open source has been a wellspring of creativity and the software it has generated has created an immense amount of innovation and value for us all. But AI is a totally different beast. What will it do? What will people do with it? Nobody knows. All we can say is – watch this space.
Sources include: Axios
In another open source and AI story, Databricks, an analytics company that has already made their AI model open source, has announced the acquisition of MosaicML, an open-source startup and competitor to OpenAI, for a whopping $1.3 billion.
MosaicML, known for its expertise in neural networks, has developed a platform that enables organizations to train large language models and deploy generative AI tools at an amazingly low cost.
Despite a previous investor-round valuation of just $222 million, Databricks was willing to pay almost six times that amount for this acquisition, perhaps because they get the entire team and not just the software.
MosaicML’s value proposition is clear; its open-source approach and focus on enabling organizations to build their own large language models using their own data have set it apart in the field.
The acquisition will see MosaicML integrated into the Databricks Lakehouse Platform, enhancing its existing multi-cloud offerings with generative AI tooling. MosaicML’s latest release, MPT-30B, has demonstrated how organizations can quickly and cost-effectively build and train their own state-of-the-art models.
Ali Ghodsi, Databricks’ co-founder and CEO, expressed excitement about the acquisition, stating that the shared vision of Databricks and MosaicML will democratize AI and make the Lakehouse the best place to build generative AI and large language models.
And as noted, the entire MosaicML team, including CEO and co-founder Naveen Rao, will join Databricks post-acquisition, taking Databricks from early entry to a relative AI powerhouse.
Sources include: TechCrunch
On the dark side of open source stories, and a move that has sent shockwaves through the open-source community, Red Hat has decided to restrict access to the source code of Red Hat Enterprise Linux (RHEL). From now on, only paying customers will be able to access the source code, and they won’t be able to legally share it.
This decision has implications for downstream projects that rebuild the RHEL source code to produce compatible distributions, such as AlmaLinux, Rocky Linux, EuroLinux, and Oracle Unbreakable Linux. These projects rely on the RHEL source code to ensure perfect compatibility with current RHEL apps, a feature that won’t be available with CentOS Stream, the only source code that Red Hat will continue to publish publicly.
This move is a blow to the open-source community, with some users expressing feelings of betrayal and violation of the General Public License or GPL that governs Linux licensing. However, Red Hat is acting within the terms of the GPL, which only requires them to make source code available to people using the binaries built from them, i.e., their paying customers.
While it’s still possible for downstream distros to obtain the source code at the point of a new major release, they won’t be able to get usable source code for each subsequent point release and the ongoing updates. This move will make the continuation of these projects substantially more difficult.
This decision by Red Hat could be seen as a continuation of the company’s strategy since it brought CentOS, the Red Hat open source clone in-house back in 2014. The move may boost IBM’s bottom line, but it risks turning public opinion against the company, which has allowed clones and rebuilds of its operating systems since its very early days.
Sources include: The Register
OpenAI, that gave us ChatGPT, is planning to create a marketplace for AI models, similar to an app store. This platform will allow developers to sell their models built on AI technology, potentially bringing the global market onto a single platform and sparking the next AI revolution.
There are already an amazing number of apps developed which use the OpenAI API – you can get them with the paid version of Open AI but it’s largely an “ad hoc” and thrown together listing of free extensions.
AI advisor Vin Vashishta sees gold in this marketplace as the next step in the Generative AI, comparing its potential impact to that of Apple’s App Store. The marketplace could democratize the AI space, opening the playing field to all kinds of AI developers and reducing any form of monopoly – and did we mention “cash cow.”
After Apple, who might have missed the boat on AI itself, still will collect untold millions from apps in its app store including the OpenAI app.
This marketplace model could disrupt the AI landscape, creating opportunities for new entrants and pushing existing players to innovate and differentiate themselves.
While OpenAI is the first to plan an app store format for all kinds of AI models, Amazon entered the field of foundational models in the marketplace two months ago with Bedrock. However, OpenAI has a leg up on others right now – it has the trust, the brand and of course, their API.
The only uncertainty lies in the monetary structuring. And OpenAI has been rather closed about disclosing any information on how they would charge the customers or developers or how they would share revenue.
Sources include: Analytics India Magazine
JP Morgan has been fined $4 million by the U.S. Securities and Exchange Commission (SEC) for deleting approximately 47 million electronic communications records from 2018 related to its Chase Bank subsidiary. These records, which were required to be retained under the Securities Exchange Act, were permanently deleted, hindering the SEC and others in their investigations.
The issue originated from a project aimed at deleting older communications and documents no longer required to be retained. However, due to “glitches,” the deletion process failed. In an attempt to troubleshoot the issue, workers deleted electronic communications from the first quarter of 2018, believing that safeguards were in place to prevent the deletion of any records required to be retained.
JP Morgan blames an unnamed archiving vendor for the mishap, claiming that the vendor had assured them that their media storage complied with the relevant Exchange Act rules regarding the 36-month retention period. However, the vendor had failed to properly apply the retention setting to the “Chase” domain within JP Morgan, leading to the permanent deletion of all emails within it, except those protected by the extra coding on “legal holds.”
In response to the incident, JP Morgan has implemented its own 36-month retention coding and overhauled its operating procedures. The SEC found that JP Morgan had “wilfully violated Section 17(a) of the Exchange Act and Rule 17a-4(b)(4) thereunder” and ordered the company to cease and desist from committing or causing any future violations, in addition to paying a penalty of $4 million.
This might lead to a new phrase – the agony of delete.
Sources include: The Register
In a new development that’s raising questions, one of which is “why?” A company named Throwflame is preparing to launch the Thermonator, a robot dog equipped with a flamethrower. The Thermonator is built on the Unitree Go1 quadrupled robot which includes a variety of cameras and sensors for autonomous navigation – it can move around like a – well, like a dog.
But the Thermonator is equipped with Throwflame’s ARC Flamethrower, capable of blasting a 30-foot long stream of burning fuel for up to 45 minutes. The good news is that the robot weighs around 60 pounds and after the flamethrower upgrade, it has limits on its operational time before needing a battery swap.
The Thermonator isn’t available for purchase just yet, interested parties can join a waitlist on Throwflame’s website, with shipping expected in the third quarter of 2023. So if you have money burning a hole in your pocket. The cost hasn’t been revealed, but the Unitree Go1 sells for between $2,700 and $3,500, and the ARC Flamethrower goes for between $699 and $899, giving a rough estimate of the Thermonator’s price.
And as the original story in Gizmodo stated, “while Thermonator may seem like an exciting piece of tech, potential buyers should consider the implications of owning a robot dog with a flamethrower, including the need for additional health, property, and home insurance.
And if you leave it at home, you have to remember to tell it, “I’ll be back.”
Sources include: Gizmodo
That’s the top tech news stories for today.
Links to all of the stories can be found in the text version of this podcast at itworldcanada.com/podcasts
Hashtag Trending goes to air five days a week with a special weekend interview episode called Hashtag Trending, the Weekend Edition. You can follow us on Google, Apple, Spotify or wherever you get your podcasts.
We’re also on YouTube five days a week with a video newscast only there we are called Tech News Day. Why don’t you check us out? Give us a like?
We love to hear your comments. You can find me on Linked In, Twitter or on our Mastodon site technews.social where I’m @therealjimlove.
Or if that’s too much to remember, just go to the article at itworldcanada.com/podcasts and you’ll find a text version with additional links and references. Click on the x or the check mark and tell us what you think.
And we read every one, and respond to them all. Even if they are — flames. We’ll just use what you say so we don’t go to the dogs. Robot dogs, that is.
Oh, one more thing. Have a terrific Tuesday!