Experts expect some inference to start moving from specialist graphics-processing units ( GPU s), which are Nvidia’s forte, to general-purpose central processing units ( CPU s) like those used in laptops and smartphones, which are AMD ’s and Intel’s.

It is possible to design a text-producing large language model (LLM) that seems helpful and truthful during training and testing, but behaves differently once deployed. And according to a study shared this month on arXiv1, attempts to detect and remove such two-faced behaviour are often useless — and can even make the models better at hiding their true nature.The finding that trying to retrain deceptive LLMs can make the situation worse “was something that was particularly surprising to us … and potentially scary”, says co-author Evan Hubinger, a computer scientist at Anthropic, an AI start-up company in San Francisco, California.

including by scaling up the use of open science, affordable and open-source technology, research and development

We reaffirmthat the creation, development and diffusion of innovations and new technologies and associated know-how, including the transfer of technology on mutually agreed terms, are powerful drivers of economic growth and sustainable development. We reiterate the need to accelerate the transfer of environmentally sound technologies to developing countries on favourableterms, including on concessional and preferential terms, as mutually agreed, andwenote the importance of facilitating access to and sharing accessible and assistive technologies

commit to exploring measures to address therisks involved in biotechnology and human enhancement technologies applied to the military domain

developingnorms, rules and principles on the design, development and use of military applications of artificial intelligence through a multilateral process

strengtheningoversight mechanisms for the use of data-driven technology, including artificial intelligence, to support the maintenance of international peace and security

commit to concluding without delaya legally binding instrument to prohibit lethal autonomous weaponssystems that function without human control or oversight, and which cannot be used in compliance with international humanitarian law, and to regulate all other types of autonomous weapons systems

international norms, rules and principles to address threats to space systems and, on that basis, launch negotiations on a treaty to ensure peace, security and the prevention of an arms race in outer space

We acknowledge that the accelerating pace of technological change necessitates ongoing assessment and holistic understanding of new and emerging developments in science and technology impactinginternational peace and security, including through misuse by non-State actors, including for terrorism

including through the transfer of technology on mutually agreed terms to help close the digital and innovation divide.

summary of Russell’s argument—which claims that with advanced AI, “you getexactly what you ask for, not what you want”

70% of respondents thought AI safety research should be prioritized more than it currently is.

etween 41.2% and 51.4% of respondents estimated a greater than 10% chance of humanextinction or severe disempowerment

Amount of concern potential scenarios deserve, organized from most to least extreme concern

ven among net optimists,nearly half (48.4%) gave at least 5% credence to extremely bad outcomes, and among net pessimists, more than half(58.6%) gave at least 5% to extremely good outcomes. The broad variance in credence in catastrophic scenarios showsthere isn’t strong evidence understood by all experts that this kind of outcome is certain or implausible

scenarios worthy of most concern were: spread of false information e.g. deepfakes (86%), manipulation oflarge-scale public opinion trends (79%), AI letting dangerous groups make powerful tools (e.g. engineered viruses)(73%), authoritarian rulers using AI to control their populations (73%), and AI systems worsening economic inequalityby disproportionately benefiting certain individuals (71%).

Most respondents considered it unlikely that users of AI systems in 2028 will be able to know the truereasons for the AI systems’ choices, with only 20% giving it better than even odds.

Answers reflected substantial uncertainty and disagreement among participants. No trait attracted near-unanimity on anyprobability, and no more than 55% of respondents answered “very likely” or “very unlikely” about any trait.

Only one trait had a median answer below ‘even chance’: “Take actions to attain power.” While there wasno consensus even on this trait, it’s notable that it was deemed least likely, because it is arguably the most sinister, beingkey to an argument for extinction-level danger from AI

‘intelligence explosion,’

he top five most-suggested categories were: “Computerand Mathematical” (91 write-in answers in this category), “Life, Physical, and Social Science” (77 answers), “HealthcarePractitioners and Technical” (56), “Management” (49), and “Arts, Design, Entertainment, Sports, and Media”

predicted a 50% chance of FAOL by 2116, down 48 years from 2164 in the 2022 survey

ay an occupation becomes fully automatable when unaided machines can accomplish it betterand more cheaply than human workers. Ignore aspects of occupations for which being a human isintrinsically advantageous, e.g. being accepted as a jury member.Think feasibility, not adoption.[. . . ]Say we have reached ‘full automation of labor’ when all occupations are fully automatable. That is,when for any occupation, machines could be built to carry out the task better and more cheaply thanhuman workers

the chance of unaided machines outperforming humans in every possible task was estimated at 10%by 2027, and 50% by 2047.

High-level machine intelligence (HLMI) is achieved when unaided machines can accomplish everytask better and more cheaply than human workers. Ignore aspects of tasks for which being a humanis intrinsically advantageous, e.g. being accepted as a jury member.Think feasibility, not adoption

High-Level Machine Intelligence’

six tasks expected to take longer than ten years were: “After spending time in a virtual world, output the differentialequations governing that world in symbolic form” (12 years), “Physically install the electrical wiring in a new home”(17 years), “Research and write” (19 years) or “Replicate” (12 years) “a high-quality ML paper,” “Prove mathematicaltheorems that are publishable in top mathematics journals today” (22 years), and solving “long-standing unsolvedproblems in mathematics” such as a Millennium Prize problem (27 years)

2,778 AI researchers who had published peer-reviewed research in the prior year in six topAI venues (NeurIPS, ICML, ICLR, AAAI, IJCAI, JMLR). This to our knowledge constitutes the largest survey of AIresearchers to date

hey are experts in AI research, not AI forecasting and might thus lack generic forecasting skills andexperience, or expertise in non-technical factors that influence the trajectory of AI.

lack of apparent consensus among AI experts on the future of AI [

was disagreement about whether faster or slower AI progress would be better for thefuture of humanity.

substantial” or “extreme” concern is warranted about six different AI-relatedscenarios, including spread of false information, authoritarian population control, and worsenedinequality.

While68.3% thought good outcomes from superhuman AI are more likely than bad, of these net optimists48% gave at least a 5% chance of extremely bad outcomes such as human extinction, and 59% of netpessimists gave 5% or more to extremelygoodoutcomes. Between 37.8% and 51.4% of respondentsgave at least a 10% chance to advanced AI leading to outcomes as bad as human extinction.

the chance of all humanoccupations becoming fully automatable was forecast to reach 10% by 2037, and 50% as late as 2116(compared to 2164 in the 2022 survey).

In this classroom, each student should not only have the opportunity to practice but also actively participate in discussions, creating an inclusive and deeply participatory learning environment.

Teaching others is a more powerful learning technique than re-reading or summarizing

ssues that keep teams from fulfilling their potential, as process loss and these include social loafing (when individuals put forth less effort when working in a group) and groupthink (when group members’ desire for conformity can lead to bad decisions)

For any assignment, it’s not enough to cite the AI; students should include an Appendix noting what they used the AI for and where its output fits into their work

you should try it out a number of times for your specific topic or concept and see how it reacts

ecause the AI can “get it wrong” students need to be aware of those limits and discussing this inclass is oneway to highlight its limitations

the tutor's value is not merely subject knowledge, but also their capacity to prompt the student to make an effort, pay close attention to the material, make sense of new concepts, andconnect what they know with new knowledge

utoring involves small group or one-on-one sessions with a tutor focusing on skills building.

hat feedback should be considered critically, and students should be asked to articulate how and why the feedback they received is effective (or not).

while the feedback may be helpful it should be coupled with an in-class discussion and clear guidelines

feed-up, feedback, and feed-forward. Feed-up serves to clearly articulatethe goals and expectations students are to meet. Feedbackreflects students' current progress and pinpoints areas requiring further development; it provides actionable advice, helpingstudents to achievetheir goals. Feed-forward helps teachers plan and tweak their lessons based on student work

they aremost common when asking for quotes, sources, citations, or other detailed information

national unification programme for 2030

ovators and th

significant model public releases within scope

introduced after the watermarking system is developed

reat unreleased AI model weights for models in scope as core intellectual property for their business,

establishor joina forum or mechanism through which they can develop, advance, and adopt shared standards and best practices for frontier AI safety,

Companies commit to advancing this area of research, and to developing a multi-faceted, specialized, and detailed red-teaming regime, including drawing on independent domain experts, for all major public releases of new models within scope

Where commitments mention particular models, they apply only to generative models that are overall more powerful than the current industry frontier (

hese voluntary commitmentsto remain in effect until regulations covering substantially the same issues come into force

designed to advancea generative AI legal and policy regime

Realizing the promise and minimizing the risk of AI will require new laws, rules, oversight, and enforcement.

only a first step in developing and enforcing binding obligations to ensure safety, security, and trus

the Administration will work with allies and partners to establish a strong international framework to govern the development and use of AI.

robust technical mechanisms to ensure that users know when content is AI generated

will be carried out in part by independent experts

immediately

ollaborative could acquire or develop and then distribute AI systems to address these gaps, poolingresources from member states and international development programs, working with frontier AIlabs to provide appropriate technology, and partnering with local businesses, NGOs, and beneficiarygovernments to better understand technological needs and overcome barriers to use

cquiresand distributes AI systems

institutions taking on the role of several of the models above

ssed by offering safety researchers withinAI labs dual appointments or advisory roles in the Project,

it diverts safety research away from the sites of frontier AI development.

such an effort would likely be spearheaded byfrontier risk-conscious actors like the US and UK governments, AGI labs and civil society groups

ccelerate AI safety researchby increasing its scale, resourcing and coordination, thereby expanding the ways in which AI canbe safely deployed, and mitigating risks stemming from powerful general purpose capabilities

xceptional leaders and governancestructures

an institution with significant compute, engineering capacity and access tomodels (obtained via agreements with leading AI developers), and would recruit the world’s leadingexperts

n AI Safety Project need not be, and should beorganized to benefit from the AI Safety expertise in civil society and the private sector

like ITER and CERN.

conduct technical AI safety research26at an ambitious scale

y to exclude from participation states who are likely to want to useAI technology in non-peaceful ways, or make participation in a governance regime the preconditionfor membership

consistently implement the necessary controls to manage frontiersystem

he Collaborative to have a clear mandateand purpose

diffusing dangerous AI technologies around theworld, if the most powerful AI systems are general purpose, dual-use, and proliferate easily.

would need to significantly promote access to the benefits of advanced AI (objective1), or put control of cutting-edge AI technology in the hands of a broad coalition (objective 2)

he resourcesrequired to overcome these obstacles is likely to be substantial,

nderstanding the needs of membercountries, building absorptive capacity through education and infrastructure, and supporting thedevelopment of a local commercial ecosystem to make use of the technology

being deployed for “protective” purposes

ncrease global resilience to misusedor misaligned AI systems

existence of a technologicallyempowered neutral coalition may also mitigate the destabilizing effects of an AI race betweenstates

access to its safe technology could be offered as an incentivefor countries to participate in a governance regime that enforces responsibility (such as agreementsto enact stricter regulation, or restrict military AI development).

reduce geopolitical instability amidst fierce AI competition among states

ensure the benefits ofcutting-edge AI reach groups that are otherwise underserved by AI developmen

legitimate international access to advanced AI

a Frontier AICollaborative could take the form of an international private-public partnership that leverages existingtechnology and capacity in industry, for example by contracting access to or funding innovation inappropriate AI technology from frontier AI developers.

ligned countries may seek to form governance clubs, asthey have in other domains. This facilitates decision-making, but may make it harder to enlist othercountries later in the process

its effectiveness will depend on its membership, governance and standard-setting processes

Governance Organization should focus primarily on advanced AI systems thatpose significant global risks, but it will be difficult in practice to decide on the nature and sophisti-cation of AI tools that should be broadly available and uncontrolled versus the set of systems thatshould be subject to national or international governance

Automated (even AI-enabled) monitoring

some states may be especially reluctant to join due to fear of clandestinenoncompliance by other states

Other AI-specific incentives for participation include conditioning on participationaccess to AI technology (possibly from a Frontier AI Collaborative) or computing resources.22Statesmight also adopt import restrictions on AI from countries that are not certified by a GovernanceOrganization

while urgent risks may need to be addressed at first by smallergroups of frontier AI states, or aligned states with relevant expertise

The impact of a Governance Organization depends on states adoptingits standards and/or agreeing to monitoring.

standard setting (especially in an international and multistakeholder context) tends to be a slowprocess

detection and inspections of large data centers

self-reporting of compliance with international standards

Where states have incentives to undercut each other'sregulatory commitments, international institutions may be needed to support and incentivize bestpractices. That may require monitoring standards compliance.

therefore enable greater access to advanced AI

international standard setting wouldreduce cross-border frictions due to differing domestic regulatory regimes

uture regulations will limit access to powerful AI technologies in jurisdictions with inadequate AIgovernanc

he InternationalTelecommunication Union (ITU)

Advanced AI Governance Organization.

intergovernmental or multi-stakeholder

as less institutionalizedand politically authoritative scientific advisory panels on advanced AI

epresentation may trade off against a Commission’s ability to overcomescientific challenges and generate meaningful consensus

broad geographic representation in the main decisionmaking bodies, and a predominance of scientificexperts in working groups

he scientific understanding of the impacts of AI should ideally beseen as a universal good and not be politicized

foundational “Conceptual Framework

Commission might undertake activities that draw andfacilitate greater scientific attention, such as organizing conferences and workshops and publishingresearch agenda

o increase chances of success, aCommission should foreground scientific rigor and the selection of highly competent AI expertswho work at the cutting edge of technological development and who can continually interpret theever-changing technological and risk landscape

internationally representative group of experts

there is significant disagreement even among experts about the different opportunities andchallenges created by advanced AI

International consensus on the opportunities and risks from advanced AI

the Commission on Frontier AI could facilitatescientific consensus by convening experts to conduct rigorous and comprehensive assessments ofkey AI topics, such as interventions to unlock AI’s potential for sustainable development, the effectsof AI regulation on innovation, the distribution of benefits, and possible dual-use capabilities fromadvanced systems and how they ought to be managed

intergovernmental body

dangerous inputs: computing resources have been targeted by US, Japanese and Dutch export controlsthat prevent the sale of certain AI chips and semiconductor manufacturing equipment to Chin

norms and standards

standards

Establish guidelines

through education, infrastructure, andsupport of the local commercial ecosystem

develop frontier AI collectively or distribute and enable access

developing and/or enabling safe forms of access to AI.

ontrolling AIinputs

tobuild consensus on risksand how they can be mitigated,andset safety norms and standardsandsupport their implementationto help developers and regulators with responsible developmentand use. International efforts toconduct or support AI safety research

national regulation may be ineffective for managing the risks of AI even within states. States willinevitably be impacted by the development of such capabilities in other jurisdictions

building capacity to benefit from AI through education, infrastructure, andlocal commercial ecosystems

build consensus on AI opportunities

Standards

SO/IEC

whether these institutions should be new or evolutions of existing organizations, whether the con-ditions under which these institutions are likely to be most impactful will obtain,

A failure to coordinate or harmonizeregulation may also slow innovation.

My worst fears are that we cause significant, we, the field, the technology, the industry cause significant harm to the world. I think that could happen in a lot of different ways. It’s why we started the company

worry that as the models get better and better the users can have sort of less and less of their own discriminating thought process around it. But, but I think users are more capable than we give often, give them credit for in, in conversations like this.

but also the genome project

to us as the American people to write the answer

t’s really like the invention of the internet in scale, at least, and potentially far, far more significant than that

in the text such as multistakeholders and networked-multilateralism

European space officials like the Artemis program and are seeking areas for greater involvement. This is drawing them closer to NASA.

However, the more significant reason is probably a political one

anuary, Josef Aschbacher, director general of the European Space Agency, said his focus remains on the International Space Station Partnership with NASA, Russia, Canada, and Japan. "For the moment we have neither the budgetary nor the political, let’s say, green light or intention to engage in a second space station—that is participating on the Chinese space station,"

Nearly six years ago the European Space Agency surprised its longtime spaceflight partners at NASA, as well as diplomatic officials at the White House, with an announcement that some of its astronauts were training alongside Chinese astronauts. The goal was to send European astronauts to China's Tiangong space station by 2022.

A new study argues that the boundary between Earth's atmosphere and outer space—known as the Kármán line—is 20 kilometers, or about 20%, closer than scientists thought

Until now, most scientists have said that outer space is 100 kilometers away.

World Air Sports Federation (FAI) in Lausanne, Switzerland, the keeper of outer space records

A country’s sovereign airspace theoretically extends to the (disputed) boundary with space.

More serious challenges for balloons might be regulatory and political.

This kind of technology could eventually be a threat—or a complement—to existing space data businesses.

The benefits include higher resolution imagery: World View’s investor presentation claims sensor resolutions of 3 to 5 centimeters per pixel, compared to the 50 cm per pixel resolution that is top of line in space-based Earth observation. The balloons also offer more persistent monitoring by floating above one location for weeks at a time, whereas Earth observation satellites can typically gather data a few times a day at most. World View already has a partnership with Maxar, a leading satellite firm with an extensive remote-sensing business.

Axiom, a space company that flies passengers to the International Space Station and is developing its own space station, said this week that most of the demand for passenger services is coming from governments without their own space programs, not tourists with deep pockets.

the project is still on the drawing board, with hopes to finalize its design by 2025. The goal is to fly the vehicle in 2027—but, as with most space technology forecasts, that’s an optimistic projection.

The US has only put a nuclear reactor in space once before, through an Air Force program called SNAPSHOT.

But the dream of space fanatics is a proper nuclear rocket, one using a fission reactor to run an engine two to three times more powerful than any motor dependent on combusting fossil fuels. Launched into space on a conventional rocket, it could shorten trips to Mars or give the Space Force unprecedented maneuverability. Last week, NASA and DARPA, the US military’s advanced tech lab, announced a collaborative project called DRACO to build and test exactly such a vehicle.

Deep space missions regularly rely on nuclear power, using the heat emitted by radioactive substances to generate electricity without an atom-splitting chain reaction. NASA’s Perseverance rover, currently exploring the surface of Mars, is powered by one of these devices.

Asteroid miners prepare for launch. AstroForge, which wants to pluck platinum-group metals from near-Earth objects, plans to launch a payload in April that will demonstrate the ability to mine metals in space. The company also expects to launch another spacecraft onboard Intuitive Machines’ lunar mission later this year that will fly to a target asteroid and assess its viability for future mining.

training for women, with a special focus on digital expertise.

The focus is to be increasingly on software solutions (digital health),

stimulating the creation of jobs offering decent working conditions. It fo-cuses in particular on the promising industries of the future, such as information and communica-tion technologies (ICT),

Support digitalisation of the African economyThe BMZ aims to effectively support the rapidly developing digital economy, for example through the Make-IT in Africa initiative, the establishment of digicentres and activities to assist African initiatives including the Smart Africa Alliance. It helps African partner countries to enhance the economic and political framework for digital transformation, to create digital markets, provide secure, universal internet access and bridge the “digital divide” within the population. It is also fostering legal standards and data privacy regulations, for example through Team Europe Initiatives such as the African European Digital Innovation Bridge Network and the EU-AU Data Flagship

digital infrastructure and health infrastructure, the BMZ aims to mobilise invest-ment –

Support the AfCFTA and ensure trade agreements are pro-development

German development policy faces the challenge of finding differentiated and flexible responses that take account of the fact that African states have their own interests, and that each state has its own view of the world and its own vision of the best economic, political and social order.

digital transformation

Pillar 2: Human Capital Development

Pillar 1: Digital Economy and Infrastructure

DTA will foster an inclusive and resilient African digital ecosystem, led by African communities and built on an open, interoperable, reliable, and secure internet

invest over $350 million and facilitate over $450 million in financing for Africa

That seems likely to change in the next few weeks, when an uncrewed lander becomes the first commercial vehicle to touch down on the Moon

The Outer Space Treaty of 1967, space law’s foundational text, is showing its age. It dates back to the era when only governments had access to space. And it states that no claims of sovereignty can be made, on the Moon or elsewhere. Efforts to update the treaty to establish rules around resource extraction have run into the lunar regolith. America has refused to sign the Moon Agreement, adopted by 18 countries in 1984, whereas China and Russia have rejected America’s latest proposal, the Artemis accords of 2020.

Peregrine lander built by Astrobotic Technology, in Pittsburgh, Pennsylvania, also operates under the CLPS programme

“commercial lunar payload services” (CLPS) programme

Nova-C, created by Intuitive Machines, a startup in Houston, Texas

One vehicle, HAKUTO-R Mission 1, operated by ispace, a Japanese firm, is already on its way and is scheduled to land in late April

Of 178 successful missions in 2022, 90 were by companies (in many cases subcontracted by governments), and of those 61 were by one firm, SpaceX.

growing support for a ban on anti-satellite weapons tests as one sign of progress, but that effort came after a series of particularly messy orbital tests, and can be seen as an effort to limit Chinese and Russian weapons development.

What might change that, the report suggests, is a disaster

most influential actors in the space economy, whether nation-states or companies, are still happier with a free hand than an insurance policy

global governance: Rules for space traffic management, protocols for space debris mitigation and removal, and norms for economic activity in space, from resource extraction to property rights.

OpenAI is developing a tool for “statistically watermarking the outputs of a text [AI system].” Whenever a system — say, ChatGPT — generates text, the tool would embed an “unnoticeable secret signal” indicating where the text came from.

he expressed the belief that, if OpenAI can demonstrate that watermarking works and doesn’t impact the quality of the generated text, it has the potential to become an industry standard.Not everyone agrees. As Devadas points out, the tool needs a key, meaning it can’t be completely open source — potentially limiting its adoption to organizations that agree to partner with OpenAI. (If the key were to be made public, anyone could deduce the pattern behind the watermarks, defeating their purpose.)But it might not be so far-fetched. A representative for Quora said the company would be interested in using such a system, and it likely wouldn’t be the only one.

“If [it] becomes a free-for-all, then a lot of the safety measures do become harder, and might even be impossible, at least without government regulation,”

Aaronson acknowledged the scheme would only really work in a world where companies like OpenAI are ahead in scaling up state-of-the-art systems — and they all agree to be responsible players. Even if OpenAI were to share the watermarking tool with other text-generating system providers, like Cohere and AI21Labs, this wouldn’t prevent others from choosing not to use it.

Unaffiliated academics and industry experts, however, shared mixed opinions.

OpenAI also declined, saying only that watermarking is among several “provenance techniques” it’s exploring to detect outputs generated by AI.

Watermarking AI-generated text isn’t a new idea

OpenAI’s watermarking tool acts like a “wrapper” over existing text-generating systems, Aaronson said during the lecture, leveraging a cryptographic function running at the server level to “pseudorandomly” select the next token. In theory, text generated by the system would still look random to you or I, but anyone possessing the “key” to the cryptographic function would be able to uncover a watermark.

“We want it to be much harder to take [an AI system’s] output and pass it off as if it came from a human,” Aaronson said in his remarks. “This could be helpful for preventing academic plagiarism, obviously, but also, for example, mass generation of propaganda — you know, spamming every blog with seemingly on-topic comments supporting Russia’s invasion of Ukraine without even a building full of trolls in Moscow. Or impersonating someone’s writing style in order to incriminate them.”

the hope is to build it into future OpenAI-developed systems

why it is working on a way to “watermark” AI-generated content

will fund musical curricula focused on conflict resolution and foreign exchange programs for young musicians across the globe

Alongside funding for military jets and naval warships, the new $857.9 billion US defense spending bill includes a program to fund musical exchanges around the globe. Dubbed the PEACE Through Music Diplomacy Act, the legislation funds US State Department cultural exchange projects that encourage artistic collaboration across borders.

People using the stimulator and their physicians could no longer access the proprietary software needed to recalibrate the device and maintain its effectiveness.